Santosh Brahmachary
Apex, North Carolina, United States
6K followers
500+ connections
About
Currently ...
Seasoned Security Leader| Principal Architect with 24+ yrs. of…
Activity
6K followers
-
Santosh Brahmachary reposted thisSantosh Brahmachary reposted thisMost GPU engineers overestimate compute. They profile a slow system, see low throughput, and assume the answer is more FLOPs, more parallelism, or a faster accelerator. But modern GPU performance usually fails for a simpler reason: The hardware is waiting. Waiting on memory. Waiting on communication. Waiting on dependency chains. That’s why “high utilization” can coexist with disappointing performance. Because utilization is not the same as useful work. A GPU is not fundamentally a compute machine. It is a machine designed to keep compute busy while the real bottlenecks stall everything else. And most CUDA intuition breaks down exactly when that distinction starts to matter. That gap between writing kernels and understanding why they behave the way they do is what pushed me to write CUDA Mastery 2026. A deep reference on the layers most resources stop before: SM behavior, memory systems, scheduling, tensor cores, PTX/SASS, and real kernel execution. If GPU performance has ever felt less intuitive than it should, this is for you. https://lnkd.in/dSXjzkzv
-
Santosh Brahmachary reposted thisSantosh Brahmachary reposted thisVery excited to see the release of the Oracle Agentic Memory SDK. This python package makes it easy to use the Oracle AI Database for storing short and long term memory for AI Agents. You can use it for free with the Oracle Free image too! I spent some time this afternoon playing with it and vibe-coding a useful Engagement Tracking app to make, search and update memories to help me avoid cognitive burden when context switching on projects. Blog: https://lnkd.in/eM9vyWQT Code: https://lnkd.in/eCEGXf98 Documentation: https://lnkd.in/eN4rVeEi
-
Santosh Brahmachary reposted thisSantosh Brahmachary reposted thisOracle Deep Data Security is a new database‑native authorization framework in Oracle AI Database 26ai that enforces identity‑ and context‑aware, row/column/cell‑level access for humans and AI agents, with declarative SQL policies and secure identity propagation from external IdPs. #AI #Agentic #Database #Oracle #26ai #RAG #Data #Security #SQL https://lnkd.in/gNVPWSK3Designing Least‑Privilege for AI: A Practical Look at Oracle Deep Data SecurityDesigning Least‑Privilege for AI: A Practical Look at Oracle Deep Data Security
-
Santosh Brahmachary reposted thisSantosh Brahmachary reposted thisAn 80B model on 8GB VRAM. No quantization. Full precision. This shouldn't work. oLLM just made it possible. Here's how: 𝘛𝘩𝘦 𝘱𝘳𝘰𝘣𝘭𝘦𝘮: qwen3-next-80B needs ~190GB VRAM for baseline inference. Most engineers don't have $15K worth of GPUs sitting around. 𝘛𝘩𝘦 𝘴𝘰𝘭𝘶𝘵𝘪𝘰𝘯: oLLM offloads intelligently: → Layer weights load from SSD directly to GPU, one at a time → KV cache offloads to SSD, loads back when needed → FlashAttention-2 with online softmax (no full attention matrix) → Chunked MLP for large intermediate projections 𝘛𝘩𝘦 𝘯𝘶𝘮𝘣𝘦𝘳𝘴 (RTX 3060 Ti, 8GB): → qwen3-next-80B: 160GB weights, 50K context, ~190GB baseline → 7.5GB with oLLM → gpt-oss-20B: 13GB weights, 10K context, ~40GB baseline → 7.3GB with oLLM → gemma3-12B: 25GB weights, 50K context, ~45GB baseline → 6.7GB with oLLM → llama3-8B: 16GB weights, 100K context, ~71GB baseline → 6.6GB with oLLM 𝘛𝘩𝘦 𝘵𝘳𝘢𝘥𝘦𝘰𝘧𝘧: Speed. qwen3-next-80B runs at ~1 tok/2s. Not for real-time chat. Perfect for: → Contract and compliance analysis → Medical literature processing → Large log file analysis → Historical chat extraction 𝘛𝘩𝘦 𝘴𝘵𝘢𝘤𝘬: Built on HuggingFace Transformers and PyTorch. Works on NVIDIA, AMD, and Apple Silicon. MIT licensed. 2.4K GitHub stars. pip install --no-build-isolation ollm Full repo: https://lnkd.in/e6dP4886 What's the first model you'd run locally if VRAM wasn't the bottleneck? 👇 💾 Save this for when your next model doesn't fit in VRAM.
-
Santosh Brahmachary reposted thisSantosh Brahmachary reposted thisA few hard earned lessons from deploying LLMs on local server: 1. Context length is usually a memory problem before it becomes a quality advantage.Teams ask for 64k or 128k context, but the real constraint is often KV cache growth, not model weights. 2. “Model fits in VRAM” is the wrong deployment test. the actual question is whether the full serving stack fits model weights, KV cache, batching, concurrency, embeddings, and all the overhead that shows up under real traffic. 3. Most instability is systems related, not model related. in practice, failures are more often caused by memory pressure, fragmentation, scheduling, or poor utilization strategy than by the model itself. 4. Long context serving is an optimization discipline. Attention kernels, cache management, batching policy, and memory allocators matter just as much as parameter count. 5. The best production model is rarely the biggest one. The model that wins is usually the one that holds quality while staying stable under latency and infrastructure constraints. That is the real shift in local LLM deployment: the conversation moves from “Which model should we run?” to “Which system can we operate reliably?” For anyone going deeper into this area, these are worth reading: vLLM / PagedAttention paper for KV cache efficiency in serving: https://lnkd.in/d2jHYWtW FlashAttention: https://lnkd.in/dTgtbU29 FlashAttention-2: https://lnkd.in/dmd-d9xt Hugging Face KV cache docs/blog: https://lnkd.in/dj7xwZ6B NVIDIA CUDA memory allocator docs: https://lnkd.in/dA42727E #LLM #AIEngineering #MLOps #vLLM #FlashAttention #OnPremAI #Inference
-
Santosh Brahmachary reposted thisSantosh Brahmachary reposted thisMost people think GPUs are compute machines. They’re not. They’re latency-hiding systems. That single misunderstanding is why so much effort in AI infrastructure gets misallocated. In training, we hide communication under compute. In inference, we try to hide serial generation under parallelism. In serving, we hide memory fragmentation behind smarter allocation. Different layers. Same game. Because the real constraint is never “how much compute you have.” It’s what’s starving that compute: – memory bandwidth – communication – or sequential dependencies If one of these dominates, your trillion FLOPs don’t matter. The system stalls. That’s why an H100 can generate tokens at human reading speed. Not because it’s weak, but because it’s waiting. Waiting on memory. Waiting on data. Waiting on the next step in a sequence it cannot parallelize. The entire modern inference stack (batching, KV cache design, speculative decoding, scheduling) is just one thing: A set of tricks to make sure something useful is happening while something else is blocked. That’s the deeper pattern. Not optimization as a checklist, but optimization as restructuring time. And once you see systems this way, everything changes: You stop asking “how do I make this faster?” And start asking “what is idle right now, and why?”Mastering CUDA and High-Performance Computing, Part XMastering CUDA and High-Performance Computing, Part X
-
Santosh Brahmachary reposted thisSantosh Brahmachary reposted thisPerformance highlights across the Oracle Acceleron family of Compute Shapes. Oracle Acceleron SmartNIC for modern workload performance ! #OCI #Acceleron #Compute #Networking #SmartNICs https://lnkd.in/eHNEmZkYAnnouncing OCI Compute E6 Acceleron Instances for Storage-Optimized and General-Purpose WorkloadsAnnouncing OCI Compute E6 Acceleron Instances for Storage-Optimized and General-Purpose Workloads
-
Santosh Brahmachary reposted thisSantosh Brahmachary reposted thisMore engineers should know about this: NVIDIA is giving away free access to 70+ AI models via hosted APIs: You get access to models like MiniMax, GLM, Kimi, DeepSeek, GPT-OSS and others - a full mix you can swap between depending on performance or use case. It works with tools like Cursor, Zed, OpenClaw, Hermes, and common agent frameworks, so you can plug it into your existing workflow without building custom infra. Setup is simple: → Generate a key at build.nvidia.com/models → Point your base_url to "𝗶𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗲.𝗮𝗽𝗶.𝗻𝘃𝗶𝗱𝗶𝗮.𝗰𝗼𝗺/𝘃𝟭" → Add your API key and select a model (e.g. minimaxai/minimax-m2.7) Under the hood it follows an OpenAI-compatible interface, so most SDKs and wrappers work with minimal changes. It’s not completely unlimited since it runs on a free tier with some usage limits. But for prototyping, internal tools, or testing different models, it’s more than enough to build and experiment without worrying about GPUs or inference costs. -- P.S. We’re building the best AI English tutor in the world. Try it to improve your speaking - it’s 15× cheaper than a human one → GetFluently.app
-
Santosh Brahmachary reposted thisDoug Burks does it again!Santosh Brahmachary reposted thisIntroducing a new PCAP tool - OhMyPCAP OhMyPCAP is a standalone web application for analyzing PCAP files. View security alerts, browse network metadata (DNS, HTTP, TLS, flows), extract ASCII transcripts, and carve individual streams - all from a single-page UI. Thanks to Chris Sanders and Jay Hawkins for the early testing and feedback!
-
Santosh Brahmachary liked thisSantosh Brahmachary liked thisVery excited to see the release of the Oracle Agentic Memory SDK. This python package makes it easy to use the Oracle AI Database for storing short and long term memory for AI Agents. You can use it for free with the Oracle Free image too! I spent some time this afternoon playing with it and vibe-coding a useful Engagement Tracking app to make, search and update memories to help me avoid cognitive burden when context switching on projects. Blog: https://lnkd.in/eM9vyWQT Code: https://lnkd.in/eCEGXf98 Documentation: https://lnkd.in/eN4rVeEi
-
Santosh Brahmachary liked thisSantosh Brahmachary liked thisOracle Deep Data Security is a new database‑native authorization framework in Oracle AI Database 26ai that enforces identity‑ and context‑aware, row/column/cell‑level access for humans and AI agents, with declarative SQL policies and secure identity propagation from external IdPs. #AI #Agentic #Database #Oracle #26ai #RAG #Data #Security #SQL https://lnkd.in/gNVPWSK3Designing Least‑Privilege for AI: A Practical Look at Oracle Deep Data SecurityDesigning Least‑Privilege for AI: A Practical Look at Oracle Deep Data Security
-
Santosh Brahmachary liked thisSantosh Brahmachary liked thisI struggled learning Linear Algebra when I was preparing for grad school. We didn't have Linear Algebra in undergrad. The mix of algebra and geometry was not intuitive to me. I wish I had a teacher to ask questions who wouldn't judge me and wasn't on the clock. I still struggle at times, wish I could do a quick refresher. So I built something I wish I'd had earlier with help of Anthropic Claude and Google Gemini : a self-study curriculum that teaches linear algebra the way 2026 actually lets us learn. Why Linear Algebra now? Language models are linear algebra at scale. 1. Every technique you read about — embeddings, attention, fine-tuning, RAG, retrieval — lives or dies on linear-algebra fluency. Without it, you're a black-box prompt user. With it, you understand *𝘄𝗵𝘆* models work, *𝘄𝗵𝗲𝗻* they fail, and *𝗵𝗼𝘄* to extend them. 2. Anchored on two textbooks: Bretscher (computational) + Saveliev (visual/geometric). No LLM hallucinations allowed! 3. Two runnable Jupyter notebooks per chapter — NumPy for numerics, SageMath for exact symbolics (highly recommend SageMath if you haven't used it) 4. This is THE fun part! A pre-built 𝗡𝗼𝘁𝗲𝗯𝗼𝗼𝗸𝗟𝗠 for every chapter: audio overview (deep-dive podcast), quiz, mind map, slide deck, infographic — 𝗮𝗻𝗱 𝘆𝗼𝘂 𝗰𝗮𝗻 𝗰𝗵𝗮𝘁 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲 𝘀𝗼𝘂𝗿𝗰𝗲𝘀! Ask questions to NotebookLM and look at the visualizations created by NotebookLM -- aren't they just beautiful? 5. Use SageMath to see already solved problems and challenge your understanding. Didn't get the concept? Ask NotebookLM what went wrong! 6. Free, open source, MIT-licensed, remix-friendly The bet: take a rigorous classical text, wrap it in modern interactive tools, and let learners cycle through six different modes — read, listen, watch, drill, code, ask — until one of them clicks. The frontier moves fast. The math underneath does not. Invest here and it compounds for a lifetime. 🔗 https://lnkd.in/guMfWAfg Star it, fork it, open a PR when an explanation falls flat. The whole thing gets better the more eyes are on it. #LinearAlgebra #MachineLearning #LLMs #AIEducation #OpenSource
Experience
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Education
Licenses & Certifications
-
-
Oracle Cloud Infrastructure 2024 Certified Architect Professional
Oracle
IssuedCredential ID 0AC2F9E232D86993C9DBABD98D2CAC00EA3950B50F1679C796B0202CA406D86A -
Oracle Cloud Infrastructure 2024 Generative AI Certified Professional
Oracle
IssuedCredential ID 651C36B357459712523E686207F4262FA0807818978210019CC2EF1BAF5D3492 -
-
Oracle Cloud Infrastructure 2024 Certified Architect Associate
Oracle
IssuedCredential ID 651C36B357459712523E686207F4262F59A88113BEFFC491E07D54309A73E249 -
Oracle Cloud Infrastructure 2024 Certified Foundations Associate
Oracle
IssuedCredential ID 651C36B357459712523E686207F4262F407F770ED2E04E920A7FAD2E41494909 -
Oracle Cloud Infrastructure 2024 Certified Security Professional
Oracle
IssuedCredential ID 07E69B75BC2CA8BC7B88B8AD101135C7C1FE40C397EBAD946721A40A3D8CA347 -
-
Courses
-
Bachelor of Engineering
-
-
HS(10+2): Science: Physics,Mathematics,Chemistry,Biology
-
Projects
-
<Local Govt. Modernization Project> Phase V
- Present
-
<Local Govt. Modernization Project> Phase IV
-
-
<Local Govt. Modernization Project> Phase III
-
-
<Local Govt. Modernization Project> Phase II
-
Environment: Oracle Public Cloud (IaaS), Products: Oracle Identity Manager 11gR2 PS3, WebLogic Server 10.3.6,12.2.1.2 SOA 11.1.1.9, , Active Directory, OID,OVD,OAM, OMSM,OMSS,WCP,SOA/BPM, ODI,APEX,OAG,OES, ,OEM,BIP,OSB,WCC,OAC, Oracle API Manager
Oracle Process
o Strictly followed Oracle Architecture Development Process (OADP)
o Conducted workshops and created requirement specifications document for OIM
o Created MoSCoW List for requirement feasibility fit-gap… Environment: Oracle Public Cloud (IaaS), Products: Oracle Identity Manager 11gR2 PS3, WebLogic Server 10.3.6,12.2.1.2 SOA 11.1.1.9, , Active Directory, OID,OVD,OAM, OMSM,OMSS,WCP,SOA/BPM, ODI,APEX,OAG,OES, ,OEM,BIP,OSB,WCC,OAC, Oracle API Manager
Oracle Process
o Strictly followed Oracle Architecture Development Process (OADP)
o Conducted workshops and created requirement specifications document for OIM
o Created MoSCoW List for requirement feasibility fit-gap analysis
o Created Design Specification document for OIM
o Created Technical Design document for OIM
o Updated architecture diagram
Integration
o Integrated OIM OAM
o Integrated OIM OMSAS OMSM
o Integrated WCP,SOA/BPM,WCC,OSB, BIP, OEM ,Weblogic Server 12.2.1.2 with OVD ,
o Integrated OIM AD with the OIM AD connector for Authoritative source reconciliation
o Integrated OIM AD with the OIM AD Password Synchronization connector
Solution Architecture
o Designed custom Self-Registration Process integration for external users using OIM SCIM REST APIs
o Designed OIM Security Policies for external, internal users.
-
<Local Govt. Modernization Project> Phase I
-
Environment: Oracle Identity Manager 11gR2 PS3 BP05 , WebLogic Server 10.3.6.012, SOA 11.1.1.9, , Active Directory, OID 11.1.1.9,OVD 11.1.1.9,Oracle Access Manager 11gR2 PS3 BP05, OMSS,WCP,SOA/BPM, ODI,APEX,OAG,OES,OPA,OEM 12c
Oracle Process
o Strictly followed Oracle Architecture Development Process (OADP)
o Conducted workshops and created requirement specifications document for OIM
o Created MoSCoW List for requirement feasibility fit-gap analysis
o Created Design… Environment: Oracle Identity Manager 11gR2 PS3 BP05 , WebLogic Server 10.3.6.012, SOA 11.1.1.9, , Active Directory, OID 11.1.1.9,OVD 11.1.1.9,Oracle Access Manager 11gR2 PS3 BP05, OMSS,WCP,SOA/BPM, ODI,APEX,OAG,OES,OPA,OEM 12c
Oracle Process
o Strictly followed Oracle Architecture Development Process (OADP)
o Conducted workshops and created requirement specifications document for OIM
o Created MoSCoW List for requirement feasibility fit-gap analysis
o Created Design Specification document for OIM OAM OID OVD OMSS
o Created Technical Design document for OIM OAM OID OVD OMSS
o Created architecture diagrams
Integration
o Integrated OIM OAM OID OVD OMSS
o Integrated OIM AD with the OIM AD connector for Authoritative source reconciliation
o Integrated OIM AD with the OIM AD Password Synchronization connector
o Integrated OIM OID by LDAPSync with custom attributes, custom LDAP Container Rules for both user and role.
o Integrated WCP SOA/BPM APEX BI with OAM OVD
o Configured OIM notification service using UMS
Development
o Developed User & UserRole Post Process Event handlers
o Developed custom notification code, notification templates
o Developed custom scheduler for custom lookups loading and maintenance.
o Developed user type, status and job title based role membership rules
o Developed user attribute based organization membership rules
o Extended OID schema for custom attributes, custom objectclass
Solution Architecture
o Designed custom Self-Registration Process integration for external users using OIM SCIM REST APIs
o Designed OIM Security Policies for external, internal users.
Achievement: First time successfully completed OMSM OMSAS Mobility App integration without prior knowledge/training
-
<Identity Management of an American telecommunications holding company>
-
Environment: Oracle Identity Manager 11gR2 PS3, WebLogic Server 10.3.6, SOA 11.1.1.9, , Active Directory, OID,OVD,OAM, OMSS,WCP,SOA/BPM, ODI,APEX,OAG,OES,OPA
Oracle Process
o Strictly followed Oracle Architecture Development Process (OADP)
o Conducted workshops and created requirement specifications document for OIM
o Created MoSCoW List for requirement feasibility fit-gap analysis
o Created Design Specification document for OIM
o Created Technical Design document for… Environment: Oracle Identity Manager 11gR2 PS3, WebLogic Server 10.3.6, SOA 11.1.1.9, , Active Directory, OID,OVD,OAM, OMSS,WCP,SOA/BPM, ODI,APEX,OAG,OES,OPA
Oracle Process
o Strictly followed Oracle Architecture Development Process (OADP)
o Conducted workshops and created requirement specifications document for OIM
o Created MoSCoW List for requirement feasibility fit-gap analysis
o Created Design Specification document for OIM
o Created Technical Design document for OIM
o Updated architecture diagram
Integration
o Integrated OIM OAM
o Integrated OIM OMSAS OMSM
o Integrated OIM AD with the OIM AD connector for Authoritative source reconciliation
o Integrated OIM AD with the OIM AD Password Synchronization connector
o Integrated OIM OID by LDAPSync with custom attributes, custom LDAP Container Rules for users and roles.
o Configured OIM notification service using UMS
Development
o Developed user Post Process Event handler
o Developed custom notification code, notification template
o Developed custom scheduler for custom lookups loading and maintenance.
o Developed user type, status and job title based role membership rules
o Extended OID schema for custom attributes, custom object class
Solution Architecture
o Designed custom Self-Registration Process integration for external users using OIM SCIM REST APIs
o Designed OIM Security Policies for external, internal users.
o Designed Mobile Device Management & Mobile Application Management Policies -
Identity and Access Management Solution on Oracle Managed Cloud Services
-
Environment: Oracle Identity Manager 11gR2 with LDAPSync, Weblogic Server 10.3.6, SOA, 11.1.1.6, Oracle Internet Directory 11.1.1.6, Oracle Virtual Directory and Oracle Access Manager 11gR2 in Oracle Managed Cloud Services Environment
Installed and configured connector
o Installed and configured Oracle eBusiness User Management connector
o Customized the EBS Responsibility and Role reconciliation scripts
OIM Catalog Enrichment
o Developed code to enrich the catalogs data… Environment: Oracle Identity Manager 11gR2 with LDAPSync, Weblogic Server 10.3.6, SOA, 11.1.1.6, Oracle Internet Directory 11.1.1.6, Oracle Virtual Directory and Oracle Access Manager 11gR2 in Oracle Managed Cloud Services Environment
Installed and configured connector
o Installed and configured Oracle eBusiness User Management connector
o Customized the EBS Responsibility and Role reconciliation scripts
OIM Catalog Enrichment
o Developed code to enrich the catalogs data with custom tags and approvers
o Customized catalog view with UDF for auto-approval and secondary approver
Troubleshooting
o Troubleshooting OIM SOA integration issues
o Troubleshooting OIM EBS integration issues -
<Identity Governance Project in one of the old and large private bank of USA>
-
Environment: Oracle Identity Manager 11gR2 PS2, WebLogic Server 10.3.6, SOA 11.1.1.6,Workday, Active Directory, Lotus Notes, CA Top Secret System, Tivoli
Designed and developed the Role Based Access Policy
o Designed and developed birthright and model access policies
o Developed java code using OIM API to upload data from excel to create and update thousands of birthright & model access policies
Installed and configured connector
o Installed and configured out of box Top… Environment: Oracle Identity Manager 11gR2 PS2, WebLogic Server 10.3.6, SOA 11.1.1.6,Workday, Active Directory, Lotus Notes, CA Top Secret System, Tivoli
Designed and developed the Role Based Access Policy
o Designed and developed birthright and model access policies
o Developed java code using OIM API to upload data from excel to create and update thousands of birthright & model access policies
Installed and configured connector
o Installed and configured out of box Top Secret (Security)/TSS Advanced connector for provisioning and reconciliations of TSS ACID, Profiles, Groups
o Enhanced the Top Secret Connector for DSN, ALIAS, XRESOURCEs, Default Group and TSO access
o Troubleshoot and resolved performance problem with the TSS connector
o Troubleshoot and resolved CFILE loading problem with the TSS connector
Designed and developed database to database integration for workday data reconciliation
o Developed java code to fetch Workday data and store in staging table
o Developed the Oracle Stored procedures and function to prepare the staging data for incremental reconciliation to OIM.
o Designed the workday to OIM trusted reconciliation integration with incremental reconciliation feature even though workday table did not store timestamp records.
o Used the Database Application Table Connector for fetching the data from staging table to OIM.
Solution Architecture
o Analyzed and reviewed requirement .for Identity Management System
o Analyzed feasibility and integration options for Identity Management System with required target systems
o Designed, configured Oracle Identity Manager and developed code for simple to complex requirements for the Identity Management System integratin with legacy to custom application systems.
-
<Identity and Access Management Solution on premise env>
-
Environment: Oracle Identity Manager 11gR2 with LDAPSync, Weblogic Server 10.3.6, SOA, 11.1.1.6, Oracle Internet Directory 11.1.1.6, Oracle Virtual Directory and Oracle Access Manager 11gR2 in Oracle Managed Cloud Services Environment
Installed and configured connector
o Installed and configured Oracle eBusiness User Management connector
o Customized the EBS Responsibility and Role reconciliation scripts
OIM Catalog Enrichment
o Developed code to enrich the catalogs data… Environment: Oracle Identity Manager 11gR2 with LDAPSync, Weblogic Server 10.3.6, SOA, 11.1.1.6, Oracle Internet Directory 11.1.1.6, Oracle Virtual Directory and Oracle Access Manager 11gR2 in Oracle Managed Cloud Services Environment
Installed and configured connector
o Installed and configured Oracle eBusiness User Management connector
o Customized the EBS Responsibility and Role reconciliation scripts
OIM Catalog Enrichment
o Developed code to enrich the catalogs data with custom tags and approvers
o Customized catalog view with UDF for auto-approval and secondary approver
Troubleshooting
o Troubleshooting OIM SOA integration issues
o Troubleshooting OIM EBS integration issues -
<Health Exchange Project on Oracle Managed Cloud Services>
-
Environment: Completed IDM suite Integration in cloud environment (Oracle on Demand). Configured components: Oracle Identity Manager 11.1.1.5 (11gR1) with LDAPSync, Weblogic Server 10.3.5, SOA, 11.1.1.5, Oracle Internet Directory 11.1.1.5, Oracle Virtual Directory 11.1.1.5, Oracle Adaptive Access Manager 11.1.1.5 and Oracle Access Manager 11.1.1.5
OIM Connector Installation and configuration
o Installed Siebel User Management connector 11.1.1.6.0
o Configured the connector for user… Environment: Completed IDM suite Integration in cloud environment (Oracle on Demand). Configured components: Oracle Identity Manager 11.1.1.5 (11gR1) with LDAPSync, Weblogic Server 10.3.5, SOA, 11.1.1.5, Oracle Internet Directory 11.1.1.5, Oracle Virtual Directory 11.1.1.5, Oracle Adaptive Access Manager 11.1.1.5 and Oracle Access Manager 11.1.1.5
OIM Connector Installation and configuration
o Installed Siebel User Management connector 11.1.1.6.0
o Configured the connector for user and employee accounts managements.
o Added custom fields. Changed Provisioning for custom fields.
Custom Java Coding for OIM
o Developed custom code using OIM 11g API for User creation, Role creation, Access Policy creation, User Role membership changes.
OID Configuration and Access Control
o Added custom attributes to user object class. Added access to users
OVD Configuration and integration with OID
o Added custom attributes. Created changelog and user adapter for OID. Created local store adapter. Created join adapter to integrate OVD to OI and AD.
Oracle Adaptive Access Manager
o Installation & Configuration of Oracle Adaptive Access Manager
o Integrated Oracle Adaptive Access Manager with Oracle Access Manager and Oracle Identity Manager.
o Implemented One Time PIN feature with OAAM
o Reset User OAAM profiles
o Implemented TAP screen authentication feature with OAAM
o Troubleshoot User login issues with OAAM
Oracle Access Manager
o Installation & Configuration of Oracle Adaptive Access Manager
o Integrated Oracle Access Manager with Oracle Adaptive Access Manager and Oracle Identity Manager.
o Added Authentication and Authorization Policies
o Troubleshoot User login issues with OAM
Troubleshooting
o Analyzed and fixed issues for IDM 11g from incorrect installation state. Corrected Metadata files. Restored OIM system from deadlock state of OIM DB.
Languages
-
English
Professional working proficiency
-
HINDI
Professional working proficiency
-
Bengali
Professional working proficiency
Organizations
-
Oracle
Oracle
- Present -
TATA Consultancy Service Limited
Associate Consultant
-
Recommendations received
9 people have recommended Santosh
Join now to viewOther similar profiles
Explore more posts
-
Kadir Tas
Harmonizing AI Guidance: Distilling Voluntary Standards and Best Practices into a Unified Framework | prepared by Center for Security and Emerging Technology (CSET) This report by the Center for Security and Emerging Technology (CSET) explores the growing need to unify global AI governance efforts through a coherent framework that consolidates existing voluntary standards, ethical guidelines, and best practices. It seeks to bridge the current fragmentation across jurisdictions and institutions, positioning standard harmonization as a prerequisite for both innovation scalability and international trust in AI systems. The study’s mission is clear: to convert the diffuse body of AI principles into a functionally interoperable, transparent, and economically efficient structure that can guide responsible adoption across sectors. Quantitatively, the report identifies over 420 distinct AI guidance frameworks currently active worldwide, spanning 65 countries and 90 standard-setting organizations. Among these, only 12% demonstrate formal cross-border compatibility, while over 40% exhibit overlapping or contradictory mandates in fairness, explainability, and accountability. The lack of harmonization, CSET estimates, contributes to global regulatory inefficiencies exceeding USD 250 billion annually by 2030. Meanwhile, organizations that adopt unified AI governance practices exhibit a 20–25% improvement in compliance efficiency and a 15% increase in trust-based market valuation**, demonstrating tangible economic incentives for alignment. The analytical framework proposed by CSET centers on three convergence pillars: Standard Mapping, which systematically links ISO, IEEE, and OECD guidelines to create a universal reference index; Policy Integration, aligning voluntary norms with enforceable legal structures such as the EU AI Act and U.S. NIST AI Risk Framework; and Operationalization, transforming abstract ethics into auditable technical and procedural benchmarks. The report further emphasizes the importance of transnational dialogue mechanisms—particularly between North America, Europe, and the Asia-Pacific—to ensure AI governance evolves as a cooperative rather than competitive process. In conclusion, Harmonizing AI Guidance defines a decisive step toward a mature global AI order—one that balances sovereignty with standardization, innovation with responsibility, and autonomy with accountability. The report asserts that harmonized frameworks are not merely regulatory tools but foundational infrastructures for the next stage of digital capitalism, where ethical alignment, interoperability, and trust will determine both economic value and geopolitical influence. #AIGovernance #CSET #AIStandards #EthicalAI #GlobalPolicy #DigitalEconomy #TechRegulation #ResponsibleInnovation #AITrust #PolicyIntegration
-
Arun Seetharaman
## The Multi-Layered Art of Content Detection in Modern DLP Strategies The evolving landscape of Data Loss Prevention (DLP) technologies, reinforces one thing: content detection has become the cornerstone of effective data security frameworks. ### Key Insights on Content Detection Technologies Detection Sophistication is Non-Negotiable: Today's DLP solutions must employ multiple detection techniques simultaneously - from RegEx patterns and dictionary-based screening to AI-powered analysis and OCR capabilities. This layered approach is essential for accurately identifying sensitive data across formats and contexts. The False Positive Challenge: One of the most significant engineering hurdles in DLP implementation remains balancing sensitivity with precision. Contextual and behavioral analytics are now critical components that help distinguish legitimate activities from potential data exfiltration attempts. Cloud-First Protection: With collaboration platforms becoming central to business operations, DLP strategies must extend seamlessly into cloud environments where data constantly moves between services, users, and devices. Adaptive Security Frameworks: The most effective DLP implementations incorporate anomaly detection and sandboxing capabilities that can identify never-before-seen threats without relying solely on signature-based detection. ### Industry Implications: What truly stands out to me is the shift toward DLP solutions that prioritize user experience alongside security. Organizations that implement heavy-handed DLP often face significant workflow disruptions and employee resistance. The most successful enterprises are taking a phased implementation approach, coupling transparent policies with ongoing user education. They're turning employees into active participants in data protection rather than seeing them merely as potential threats. ### Looking Forward: As AI continues to evolve on both sides of the security equation, we'll see increasingly sophisticated content detection capabilities that can understand context, intent, and data relationships in ways that today's systems cannot yet achieve. What's your experience with DLP implementation? Has your organization found the right balance between security, privacy, and productivity? Share your thoughts in the comments, and let's connect if you're navigating these challenges in your organization. I'm always open to discussing effective data protection strategies that don't compromise operational efficiency. #DataSecurity #DLP #ContentDetection #CyberSecurity #EnterpriseIT #DataProtection
-
ᏚᎻᎪᎠᎩ ᎬᏞᏴᎾᎠᎩ
Can we unify ISO 27001, NIST CSF 2.0, SOC 2, CIS, and others into one control set to reduce “compliance fatigue” and duplicated work? Today, most organizations operate under multiple cybersecurity & privacy frameworks: ISO 27001 for international trust, NIST CSF 2.0 for risk maturity, SOC 2 for customer assurance, CIS benchmarks for technical hardening, …and sometimes even more (PCI-DSS, HIPAA, GDPR, etc.). The outcome? Endless mapping. Endless evidence collection. Endless audits. This growing burden has a name: Compliance Fatigue. Teams spend more time managing spreadsheets and duplicate documents than improving security. But what if we treated security frameworks differently? What if we: 🔹 Built a unified control library mapped across standards 🔹 Eliminated duplicate evidence requests 🔹 Automated repeat control testing 🔹 Linked policies/procedures to all frameworks at once 🔹 Turned compliance from a checkbox into a strategic advantage The truth is: 80% of major cybersecurity standards overlap. They speak different languages, but they ask for the same foundations — governance, risk, access control, change management, asset inventory, incident response, data protection, and monitoring. A single “meta-framework” approach can: ✅ Reduce audit time ✅ Cut documentation effort ✅ Improve consistency ✅ Strengthen real security outcomes ✅ Drive faster certification cycles As GRC evolves, unifying frameworks is no longer a dream — it’s becoming a necessity. Question to the community: Have you tried building one control set to cover ISO 27001, NIST CSF 2.0, SOC 2, CIS, and more? What challenges or successes have you seen? #Cybersecurity #GRC #ISO27001 #SOC2 #NISTCSF #CIS #Compliance #RiskManagement #Governance #Audit #InformationSecurity #SecurityLeadership #ContinuousCompliance
31 Comments -
Cyber Security Times
🛡️ 𝗦𝗲𝗰𝘂𝗿𝗶𝘁𝘆 𝗢𝗽𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝘀 𝗖𝗲𝗻𝘁𝗲𝗿 (𝗦𝗢𝗖) 𝗥𝗼𝗹𝗲𝘀 𝗘𝘅𝗽𝗹𝗮𝗶𝗻𝗲𝗱 A SOC is a layered defense where every role matters: 🧭 SOC Manager → Strategy, processes, and alignment 👀 L1 Analyst → 24/7 monitoring & alert triage ⚔️ L2 Analyst → Incident response & containment 🎯 L3 Analyst → Threat hunting & advanced analysis Together, they turn alerts into action and threats into lessons. Which SOC role are you in — or aiming for next? 👇 🔔 Follow Cyber Press ® for more cybersecurity tips! #CyberSecurity #SOC #BlueTeam #IncidentResponse #ThreatHunting
17 Comments -
SECITHUB.COM | The HUB for Cyber Security, Cloud & IT Infrastructure insights
Browsers have become the biggest attack surface where compliance, visibility, and data protection all converge. A proxy security strategy is no longer optional. It’s how modern organizations enforce Zero Trust, strengthen ISO compliance alignment, and stop breaches before they happen. Discover how Proxy integrates with SASE See why unmanaged browsing costs companies millions every year Learn how to design a scalable, compliant proxy framework for hybrid teams read the full guide https://lnkd.in/dJARdp-s #Cybersecurity #ProxySecurity #SASE #ISO27001 #ZeroTrust #Compliance #CloudSecurity #NetworkSecurity #SMB #Secithub #TechLeaders #DataProtection #BrowserSecurity #EnterpriseIT #InfoSec
-
Robert W.
New Post: CISA: New Guidance for SIEM and SOAR Implementation - https://lnkd.in/d4B8q9C4 05/27/2025 10:00 AM EDT Today, CISA, in collaboration with the Australian Signals Directorate’s Australian Cyber Security Centre and other international and U.S. partners, released new guidance for organizations seeking to procure Security Information and Event Management (SIEM) and Security Orchestration, Automation, and Response (SOAR) platforms. This guidance includes the following three resources: Implementing SIEM and SOAR Platforms – Executive Guidance outlines how executives can enhance their organization’s cybersecurity framework by implementing these technologies to improve visibility into network activities, enabling swift detection and response to cyber threats. Implementing SIEM and SOAR Platforms – Practitioner Guidance focuses on how practitioners can quickly identify and respond to potential cybersecurity threats and leverage these technologies to streamline incident response processes by automating predefined actions based on detected anomalies. Priority Logs for SIEM Ingestion – Practitioner Guidance offers insights for prioritizing log ingestion into a SIEM, ensuring that critical data sources are effectively collected and analyzed to enhance threat detection and incident response capabilities tailored for organizations. CISA encourages organizations to review this guidance and implement the recommended best practices to strengthen their cybersecurity. For access to the guidance documents, please visit CISA’s SIEM and SOAR Resource page. Robert Williams#News247WorldPress
-
Supriya Ghosh
Part 5 — Operationalizing Trust in AI-Driven API Security Building human-AI synergy and ethical frameworks is only the beginning. To truly succeed, organizations must operationalize trust with measurable metrics that guide AI security maturity. 1. Transparency Scores Measure how well AI security systems explain their decisions. Track % of alerts with clear rationale shared with analysts. Include documentation of AI model changes and training data origins. Example: A healthcare system scores high when every patient record access flagged by AI comes with a human-readable explanation 2. Decision Traceability Ensure every AI security action is auditable from detection to resolution. Log decision points: who reviewed, what data was considered, final action taken. Build dashboards that show alert lifecycle and reviewer feedback. Example: An e-commerce platform can trace how a bot alert led to adaptive throttling and review, enabling compliance audits 3. Customer Perception Indices Gauge end-user trust and satisfaction related to AI-driven security. Collect feedback after security events or authentication challenges. Monitor churn or support requests tied to false positives/negatives. Example: A fintech app reduces login friction while maintaining security, reflected in a rising Net Promoter Score (NPS) 4. Anomaly Detection Effectiveness Regularly evaluate AI’s precision and recall in identifying genuine threats. Periodic reviews comparing AI alerts vs confirmed security incidents. Track Mean Time to Detect (MTTD) improvements. Example: Adaptive AI models in a cloud service reduce false positives by 30% year over year 5. Ethical Compliance Audits Maintain ongoing checks for bias, fairness, and privacy adherence. Independent reviews of AI training datasets and decision outputs. Use synthetic data tests for fairness evaluation. Example: A health tech company routinely audits AI-driven access controls to avoid demographic biases 6. Cross-Functional Collaboration Metrics Quantify team engagement and responsiveness. Measure frequency of AI-human sync meetings and action items closed. Track reduction in security incident fallout via collaboration. Example: Security and product teams at a retail platform document improved incident response times as collaboration matures. Operationalizing trust is about making AI security visible, auditable, user-centric, and accountable. By tracking these metrics, organizations embed trustworthiness into the core of every API interaction. #AITrustMetrics #SecurityMaturity #DecisionTraceability #TransparencyScores #CustomerTrust #SecurityGovernance #RiskManagement #DataEthics #SecurityMetrics #ProductSecurity
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content