Skip to content

Reliability Engineer·AI Builder·Army Veteran

Rob Salamon

I build multi-agent systems for safety-critical work.

MS Applied AI/ML·Active TS/SCI Clearance·Sacramento, CA

LinkedIn GitHub Contact Resume

Trusted on programs at

  • Lockheed Martin Space
  • Collins Aerospace
  • Oshkosh Defense
  • U.S. Army
  • Active TS/SCI Clearance
  • 10+ yrs defense engineering
  • MS Applied AI/ML
  • 487 passing tests in production

Featured Project

Agentic Reliability Engineer

A multi-agent platform that automates MIL-HDBK-217F failure-rate predictions.

See all projects
What it doesProduction

Turns raw electronics BOMs into contractual reliability deliverables in minutes — work that takes a senior engineer two weeks per board.

A Claude-powered platform with a five-stage agentic pipeline. LLMs handle judgment-heavy data work; deterministic engines handle the math. Unvalidated model output never enters the system of record.

  1. 01BOM column mapping — Claude Sonnet, confidence-gated
  2. 02Part classification — deterministic heuristic, 88 test cases
  3. 03Waterfall data resolution — cache → DigiKey → Claude Haiku vision
  4. 04Deterministic 217F calculator — pure Python, never LLM
  5. 05Artifact rendering — XLSX with live formulas, IEEE 1413 PDF, JSON twin
PythonTypeScriptNext.jsFastAPISupabaseClaude SonnetClaude Haiku
// safety boundary
agents.propose(data_extraction)
engines.decide(arrhenius_math)
// unvalidated LLM output never enters system of record
487
Passing Tests
Across calculator, classifier, and pipeline
<0.01%
Calculator Error
Validated against legacy reference data
Multi-Model Orchestration/architecture
  • Sonnet
    high-consequence reasoning
    column mapping
  • Haiku
    high-frequency vision
    $0.02 / page
  • Heuristic
    classification
    zero LLM cost

Each model chosen through empirical evaluation of cost, accuracy, and failure modes — not vibes.

~80 hrs
Manual Time Replaced
per BOM, end-to-end
$0.30
LLM Cost per BOM
Sonnet + Haiku, blended
11
Legacy Bugs Found
via 3-source data audit

Repository is private due to professional sensitivity — request reviewer access.

Recent Writing

Thinking out loud.

See all writing
PostMar 28, 2026

Shipping multi-agent systems you can actually trust

Lessons from putting a five-stage agentic pipeline into a regulated, audit-heavy workflow — and the boundary between LLM judgment and deterministic math.

1 min readRead →
PostFeb 14, 2026

Reliability is a data problem hiding in an engineering job

Sixty-year-old reliability workflows collapse the moment you treat them as a dataset. Here's what I see when I squint at a FMECA from the data side.

1 min readRead →

Research

Agentic Reliability: A Threat Taxonomy Across Five Layers of the AI Agent Stack

Technical paper · in progress

Analyzing where trust boundaries break down as agents gain autonomy. Motivated by firsthand experience shipping production agents — and watching where they actually fail.

See all research
T.01
Prompt Injection via Web Content
search layer
T.02
Poisoned Context Windows
agent harness
T.03
Malicious MCP Servers
agent harness
T.04
Compromised Training Data
operational reasoning
T.05
Permission Escalation
agent harness
T.06
Agent-to-Agent Manipulation
outward-facing audience

Framed across five layers of the agentic architecture: agent harness·search layer·web data·operational reasoning·outward-facing audience

Experience

The Arc

  1. 2023 — Present

    Lockheed Martin Space

    Reliability Engineer, Senior Staff

    Delivering reliability models and analysis for Critical Design Reviews across billion-dollar space programs. FMECA, Fault Tree Analysis, reliability growth modeling. Managing Top-Secret data sets.

  2. 2021 — 2023

    Collins Aerospace (Raytheon)

    Manager, Data & Analytics — AI/ML

    Led a cross-functional team that built a Python application cutting $10M+ in purchasing spend. Designed UI/UX, data pipelines, OCR integration. Ran Scrum sprints.

  3. 2019 — 2021

    Collins Aerospace

    Senior Reliability Engineer

    Overhauled reliability processes that hadn't changed in 60 years. Built cross-functional FMECA processes across multiple military contracts.

  4. 2016 — 2018

    Oshkosh Defense

    Senior Managing Reliability Engineer

    Created Reliability-as-a-Service ($200K year one). Led team of 7. Published analysis saving a customer $1.2M. This is where ML/AI first caught my attention.

  5. 2010 — 2017

    U.S. Army — Wisconsin National Guard

    Sergeant, Crew Chief — UH-60 Blackhawk

    Seven years maintaining and flying Blackhawk helicopters. This experience shaped everything that followed — it's why I've spent my career making defense equipment safer and more dependable for the people who depend on it.

Off-Hours

Life outside the keyboard.

The same habits that make me a decent engineer — patience, first-principles thinking, the need to understand a system all the way down — show up everywhere else in my life too.

Dad

Raising a little engineer.

Currently coaching my son's Little League T-ball team. Best part of the week.

Trails

Happiest on a mountain.

The firewatch palette on this site isn't a coincidence. Weekends I'm outside.

Garage

Bringing a project car back.

Same engineering instinct — measure, diagnose, fix, verify. Different problem surface.

House

Understanding things all the way down.

Fixing the house myself. I'd rather know how the wall is built than guess.

More on life outside work

Off the Clock

Education & Community

Education
  • 01
    MS Applied Artificial Intelligence & Machine Learning
    University of San Diego
  • 02
    BS Mechanical Engineering
    University of Wisconsin - Milwaukee
  • 03
    High School Study Abroad
    Obendorf, Germany
Community
  • 01
    Starting Block AI
    Helping small business owners in my community adopt AI tools that actually fit their workflows — not just the latest shiny thing.
  • 02
    SAE Standards Committee
    Lifecycle Cost & Logistics Supportability — voting member since 2016.
  • 03
    Mentor & Coach
    FIRST Robotics mentor (2016 — 2019). Currently coaching my son's Little League T-ball team.

Connect

Let’s build something.

I’m looking for my next role in agent systems and AI safety research. I work best with smart people solving hard problems — and I bring a decade of defense engineering, a shipped multi-agent platform, and an active security clearance to the table. Available within two weeks.

Send me a noteLinkedIn