Justin Karr

Machine Learning Engineer. I build AI agents that run in production.

20+ Years Intelligence & Data Operations · 11 AI Agents in Production

Justin Karr

About

I spent 16 years at NGA and 5 years in the Marine Corps turning raw data into actionable intelligence — three combat deployments, thousands of analyst hours automated, and a career built on making complex systems work under pressure.

Now I build AI. Not demos or prototypes — production systems with real users. I designed and operate a multi-agent platform running 30+ automated pipelines, processing thousands of records daily. I work across the full lifecycle: from identifying the problem, to prototyping with stakeholders, to hardening and deploying at scale.

I'm at my best when the problem is ambiguous and the solution needs to actually ship.

What I build

ATLAS Multi-Agent Platform

AI orchestration platform coordinating 11 specialized agents across lead intelligence, outreach automation, and real-time monitoring.

  • 30+ production pipelines running daily on a single VPS
  • Hybrid scoring system combining rule-based filters with LLM verification across 5 dimensions
  • Model routing layer with budget tracking and cost optimization
  • 12+ external API integrations, 4 databases, webhook event architecture
PythonClaude APISupabaseNotion APIFastAPIDiscord

Storm Intelligence Pipeline

Real-time severe weather monitoring system that detects hail events and triggers automated property lead collection across 11 states.

  • < 5 minute response time to NOAA severe weather alerts
  • Automated forecast → property collection → outreach pipeline
  • GeoJSON parsing with multi-state coverage
PythonNOAA APIGeoJSONAsync Processing

LEDGER Enrichment Engine

7-stage async enrichment pipeline that transforms raw business filings into scored, qualified leads with full contact profiles.

  • 5,400+ filings processed, multi-source data fusion
  • Tiered classification (A/B/C) feeding outreach prioritization
  • Browser automation with anti-detection for protected data sources
PythonPlaywrightAsync/AwaitPostgreSQL

RAG Documentation System

Production-grade retrieval system for multi-version enterprise documentation using semantic search.

  • Handles evolving datasets — add new versions without retraining
  • Metadata-aware responses with cross-version comparisons
LangChainChromaDBOpenAI EmbeddingsVector Search

Skills

AI/ML & Agents

LLM IntegrationMulti-Agent OrchestrationPrompt EngineeringHybrid ScoringRAG ArchitectureEvaluation Frameworks

Languages

Python (Expert)SQLJavaScriptR

ML Stack

PyTorchscikit-learnPandasNumPyTensorFlowLangChain

Databases

PostgreSQL/PostGISSupabaseChromaDBFAISSNotion APIOracle

APIs & Integration

FastAPIWebhook ArchitectureOAuth12+ Production Integrations

Infrastructure

Linux AdministrationCron OrchestrationGitCI/CDDocker

Experience

2025 – Present
Founder & Lead ML Engineer ATLAS AI Platform
2024 – 2025
Senior Data Scientist ECS Federal (NGA)
2022 – 2023
Senior Data Scientist Integrity First Technologies (NGA)
2019 – 2021
Business Analyst / GIS Specialist SAIC (NGA)
2004 – 2018
Intelligence Analyst & GEOINT Specialist NGA / USMC

Let's talk.

I'm open to full-time ML engineering roles, contract work, and interesting problems.

Book a Call