Kyle Palko

Nominal's partnership with HII, the world's leading manufacturer of autonomous unmanned underwater and surface vehicles, is in the wild. We're working with HII to modernize how test and mission data are collected, validated, and analyzed across their REMUS UUVs and ROMULUS USVs. Autonomy programs move at the speed of their test-and-learn loop, and the mandate from the US Navy is clear: more, faster. The upshot? During our 2025 pilot, we helped compress analysis tasks from hours to minutes and cut certain production test steps by half. Thanks to Eric Chewning and the HII team for the partnership. Read more here: https://lnkd.in/gJ-T9yyb And here: https://lnkd.in/gsJzvg8z

148

2 Comments

Earlier this week Scout AI was awarded a major contract under the US Army's Unmanned Systems (UxS) initiative, a large effort to advance ground autonomy We will integrate Fury (our autonomy platform) onto Army furnished Infantry Squad Vehicles (ISV's) The program spans a 16-month performance period, after which the Army may procure units up to $150 million in contract value Scout is collaborating exclusively with Textron Systems Corporation on this effort https://lnkd.in/ewmGMSq3

260

18 Comments

This is such an incredibly important and ongoing conversation that exists in the loose, stovepiped, & conceptual "name-your-tech-of-interest" realms of Defense. I'll cover my area of expertise, intelligence... Whether your LLM is making assessments, your object detector is espousing false confidence metrics, or your black box neural network is fusing 30 sources of multi-INT...it's all being sold as decision quality data, a force multiplier, and 'more' accurate than "that other company's" capability. 1. Decision-quality data it is not. That may be true for the commercial realm, but around here 95% "accuracy" doesn't cut it. We measure success in human lives (combatant & civilian) saved and lost on BOTH sides. 2. Force multiplication for most of today's product offering is actually analysis obfuscation. At the end of the day, whatever automated assessment 'confidence' you're bringing needs to pass the keen eyed & rigid scrutiny of an overworked & untested 17-year-old "analyst-in-the-loop." Look me in the eyes and tell me they've ever been 85% confident about anything. Now stick an easy (Send to link) button next to a polished Gen-AI assessment... Artificial Intelligence cannot be accepted as Automated Intelligence. We're forgetting the other big 'A' in this equation. AUGMENTATION. The goal for efforts like object detection began with an endstate of replacing analysts. We're not naive, that's not news. But when those efforts failed to deliver Analyst-quality outputs, we slapped an "Analyst-in-the-loop" bumper sticker to the program AND. CHANGED. NOTHING. Tell me how an analyst workflow speeds up while maintaining quality, changing from: - Look at every image and report every threat object; to - Look at every image and report every threat object...now with labeled boxes around 'most' things. Speed gained (easy button acceptance) = Tradecraft lost (due diligence) That is why I've fought so hard (and won) to bring companies like RAIC Labs (formerly Synthetaic) to analysts. Hyper-scale search and net-new threat discovery? Hell yes. Augmented target custody maintenance? Oh yeah. Automated detection reporting? Hell NO! They don't hide false positives; they accept the fact that reduction of false positives equates to loss of true positives. Instead? Cluster! It's why I continue to advocate for Unstructured: - Augmenting GenAI data shortcomings crucial for RAG enablement & structured data extraction (you know, the stuff that actually makes automation reliable). Gov LLM usage must be grounded and reliable. We must start with the foundational enablement. These folks take AUGMENTATION seriously because they started with the INTENT TO AUGMENT. Automation seeks to augment decision-makers, obfuscating analysis. Augment the analyst!! Not the decision-maker and certainly not the weapon system. Great job, Missy Cummings & NeurIPS. Whole-heartedly support your position. Thanks to Ido Banai for tagging me and instigating this rant ; )

17

3 Comments

Excited to get back to work Monday and dive deeper into semantic modeling layers. A common semantic interface can replace data translations, reduce app-to-app dependencies, simplify data sharing, and localize schema changes to applications ultimately ✨ reducing technical debt ✨. Lessons learned from Dave Wells workshop “The Data Interoperability Imperative” at Data Modeling Zone.

18

I never planned on joining a space startup. I remember when James New, PhD first talked to me about Odin Space. What struck me was how baffling the problem was. 𝗙𝗶𝗿𝘀𝘁, 𝘁𝗵𝗲 𝗴𝗮𝗽 𝘄𝗮𝘀 𝘀𝘂𝗿𝗽𝗿𝗶𝘀𝗶𝗻𝗴. We are planning missions and launching thousands of satellites, yet the basic risk data and insurance infrastructure that other major industries depend on is not where it needs to be. 𝗦𝗲𝗰𝗼𝗻𝗱, 𝘁𝗵𝗲 𝗶𝗻𝗲𝘃𝗶𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆. If the space sector continues to grow, and it will, it cannot function without accurate measurements of sub-centimeter debris, without understanding orbital risk in real time, and without a way to attribute failures when something goes wrong. You simply cannot build a resilient orbital economy without a trusted data layer that explains what is happening to your spacecraft. The more we work on this, the clearer it becomes. Someone has to build the in-orbit data foundation that allows spacecraft to be protected, insured, and failures understood. And somewhere along the way, this work began to feel much larger than a company. It is hard to ignore the sense that we are helping lay the groundwork for an industry that will shape the next century (dare I say humankind?). In its own way, it feels historic. And although the vision convinced me, I suspect a few pints helped too.

98

10 Comments

For those of you at #AIAASciTech don't miss our paper on an open source photorealistic simulation and associated tools that help explore solutions to the collision avoidance problem, an area of active research in #Autonomy. The presentation is in Bayhill 28 at 3:30 p.m. by Michael Acheson You can find our paper at https://lnkd.in/eJ8ZqFZB Abstract: Autonomous collision avoidance (ACA) for aerospace systems is an area of active study within the autonomy research community. This problem consists of three consecutive subtasks: perception, prediction, and planning. Each of these sub-tasks are large, active fields of research on their own. For this reason, this team developed an open-source simulation frame work that enables the autonomy community to target ACA using a multirotor vehicle which must avoid moving baseballs. The Baseball Avoidance Multirotor (BAM) simulation framework uses MATLAB® and Simulink® with a multirotor dynamics model in combination with the high-fidelity photo-realistic and sensor simulation capabilities of Microsoft∗ AirSim or Colosseum with Unreal® Engine. The multirotor was selected as the aerospace vehicle due to its simple dynamics and control methodologies that are accessible, high precision, and well known within academia and industry. These features allow researchers to focus on one or more ACA subtasks without managing the more challenging aspects of vehicle control. This paper presents the BAM simulation and details a series of collision avoidance challenges provided with accompanying datasets. Results from BAM simulation experiments demonstrate collision avoidance with both an algorithm that combines Optimal Reciprocal Collision Avoidance (ORCA) with Bernstein polynomials, and a transformer-based replanning method

18

🤔 Tech companies are making massive promises about how they're transforming shipbuilding. But software can't bend steel. 🏭 ⚓ At the worst moment in history for American shipbuilding, we have no time for reality distortion fields about "just doing it differently." 🛑 None of the defense tech new entrants has yet built and displayed a ship. But tech entrants bring real productivity increases to #shipbuilding. They bring fresh capital, energy, and some of America's best talent. They bring bold vision too. ⚒️ 🏗️ Together, defense tech and shipbuilding can build. We can also help fix broken parts of the Navy we love. Check out my piece in War on the Rocks ⤵️ for a discussion of these issues - Steel and Silicon: Shipbuilding's Defense Tech Moment

154

22 Comments

Shipping software isn't the goal. Shipping outcomes is. For my team, Starfox at Section 31, that means delivering capabilities to the Combined Space Operations Center (CSpOC) that enable Guardians to make faster, more informed decisions. We're not just modernizing systems; we're delivering real mission impact. Proud to be part of a team that lives this every day. Rise8 #WeShipOutcomes

29

1 Comment