Threat Intelligence & OSINT

Austin Smith

From physical reconnaissance to digital intelligence — I build production-grade OSINT pipelines, government contracting automation, and AI-powered research tools that ship, scale, and survive independent verification.

15+ OSINT Repositories
r = 0.62 Core Finding (p < 0.001)
8+ yrs Security Experience

About Me

I'm a 29-year-old from Arkansas who applies reconnaissance, threat assessment, and security skills — trained across federal, state, and private security roles — to analyze complex data and track adversarial influence online.

I swore into the Army National Guard at 17, trained at Fort Benning as a Cavalry Scout (19D), and later worked in state corrections and private security (G4S). These roles demanded strict objectivity: observing environments, documenting facts, and separating actionable threats from background noise.

In late 2024, I noticed severe polarization and coordinated narratives dominating social media. The mainstream narratives from both sides felt like coordinated influence campaigns. I approached this the way I was trained — by tracking signals, not decoding content. I began teaching myself Python in October 2025 to automate this tracking, and within months built multi-agent AI pipelines processing unstructured data at scale.

Today I ship production systems. Two live commercial products, 15+ open-source repositories, and a documented analytical methodology that survived independent stress testing. I build OSINT pipelines, procurement automation, and AI-powered research tools — and I'm available for contract work where precision, security, and reproducibility matter.

What Drives Me

🎯

Evidence Over Narrative

Every finding ships with the scripts to verify it. Failed predictions are documented publicly — negative findings are data.

🔍

Accountability & Transparency

Maintaining strict analytical objectivity — documenting correlations without claiming causation, and openly correcting past errors.

🛡️

Protecting People

From military service to OSINT research, the mission is the same — identify threats before they become harms and make information accessible to everyone.

Technical Capabilities

📡 OSINT & Threat Intelligence

  • Open-source intelligence collection & analysis
  • Temporal pattern analysis & signal detection
  • Adversarial influence tracking
  • Multi-platform narrative monitoring
  • Source triangulation & verification
  • Multi-AI cross-verification methodology

📊 Quantitative Methods

  • Pearson & Spearman correlation analysis
  • Permutation testing & bootstrap methods
  • Granger causality & lag analysis
  • K-means & hierarchical clustering
  • Chi-square & Mann-Whitney U tests
  • Causal inference (DoWhy framework)

🤖 AI & Automation

  • Multi-agent LLM pipelines (Claude, Llama Scout 17B)
  • Local entity extraction (GLiNER2) — zero API cost
  • Brave Search API integration & web-grounded verification
  • Document parsing pipelines (PDF/DOCX/OCR)
  • AI safety research & bias detection
  • Prompt engineering & verification layers

💻 Development & Data

  • Python (Pandas, SciPy, Statsmodels, NetworkX)
  • Flask web apps with CSRF, rate limiting, CSP headers
  • Chart.js dashboards & PyVis interactive graphs
  • GitHub Actions CI/CD & DigitalOcean deployment
  • Git & GitHub (15+ public repositories)
  • Dataset design, cleaning & validation

Featured Projects

Procurement Automation

Government Bid Solicitation Pipeline

An 8-stage automated bid response pipeline for small service contractors. Ingests solicitations from email or portal upload, parses PDFs/DOCX/scanned documents via OCR, extracts structured requirements (insurance, bonding, forms, pricing format), makes bid/no-bid decisions against configurable business rules, builds trade-specific pricing from templates, and generates 95%-ready bid packages. Configurable for any trade, service area, or state portal.

Pythonpdfplumberpython-docxOCRClaudeGitHub Actions
Civic Tech & AI Safety ● Live

Bill Translator

A Flask web application solving legislative transparency — simplifying dense legal jargon while preserving exact legal intent to meet Arkansas Act 602's 8th-grade readability requirement. Features custom meaning drift detection that flags potential intent shifts automatically.

PythonFlaskAnthropic APITextstatNLP
AI & Automation

The Speaker — OSINT ChatBot

A Bring Your Own Key chat interface powered by the Anthropic Messages API (Claude) with a bundled Knowledge Base. Users supply their own API key in the browser — the server never stores keys. Loads a structured OSINT knowledge base from markdown files at startup, prioritizing verified research over training data.

PythonFlaskAnthropic APICSRF ProtectionRate LimitingRender
Commercial Product ● For Sale — $50

OSINT Intelligence Pipeline

A deploy-in-15-minutes intelligence pipeline for any research topic. Ten automated stages turn three API keys into a live analytical dashboard — no servers, no databases, no DevOps required. Includes Brave Search + Claude intelligence gathering, GLiNER2 + Llama Scout dual-engine entity extraction, three-pass investigation protocol, web-grounded fact checking, rhetoric-vs-reality gap analysis, and interactive NetworkX + PyVis entity graphs. Ships as a one-time purchase with setup documentation.

PythonBrave Search APIClaudeGLiNER2Llama Scout 17BNetworkXPyVisGitHub Actions
Signal Intelligence

UVB-76 Structured Signal Analysis

15-year empirical analysis of Russia's UVB-76 shortwave station transmission timing patterns. Used K-means clustering and permutation testing (p < .0005) to identify four distinct operational modes, revealing a shift from training-cycle patterns to event-driven signaling aligned with real-world military operations.

PythonK-means ClusteringPermutation TestingTime Series
AI Safety & Online Harms

AI Manipulation OSINT Case Study

Quantitative documentation of selective emphasis in AI-assisted OSINT research. Identified a 207:1 keyword-mention disparity across AI platforms during a two-month investigation, with cross-platform verification and documented asymmetric guardrail application. Includes raw data and full methodology for independent verification.

PythonPandasCross-platform AnalysisAI Bias Detection
Geopolitical OSINT

Sovereign Capital Audit

Mapping the structural dependency of the US Defense Industrial Base on Gulf Sovereign Wealth (Mubadala/PIF) and East Asian Industry. Verified OSINT assessment tracking capital flows, supply chain vulnerabilities, and national security implications.

OSINTFinancial AnalysisSupply ChainGeopolitics
Data-Driven Policy Analysis

State Policy Analysis

Examines how U.S. states respond to the data center and energy infrastructure boom — from regulatory accommodation to moratoriums. Statistical analysis of 42 federal funding events across 27 states with transparent correction of prior methodological artifacts.

PythonSciPyDoWhyPandasStatistical Testing

Experience

Oct 2025 — Present

Independent Developer & OSINT Researcher

Self-Directed

  • Built and deployed Friction Breaker, a live commercial SaaS product with tiered subscription pricing, Discord OAuth, and a mechanism classification engine running on Claude Sonnet 4.6 + GLiNER2 local entity extraction
  • Designed and shipped an 8-stage government bid solicitation pipeline with PDF/DOCX/OCR document parsing, requirement extraction, bid/no-bid decision logic, trade-specific pricing engines, and portal submission automation
  • Built a seven-stage automated intelligence pipeline tracking 15 leverage nodes using Brave Search API, Claude, GLiNER2 (local, zero-cost), Llama Scout 17B, and NetworkX — published as a commercial product with under-15-minute setup
  • Identified and documented statistically significant correlations (r = 0.6196, p = 0.0004) between geopolitical events and institutional positioning across 8 years of data — validated by 16 independent stress tests
  • Published 15+ public repositories with reproducible findings, open datasets, and transparent methodology — including openly documented failed predictions
  • Conducted quantitative AI bias research documenting a 207:1 keyword-frequency disparity across AI platforms and asymmetric guardrail application
  • Built a legislative readability tool (Flask + Anthropic API) with custom meaning drift detection, deployed live on Render
  • Developed a BYOK OSINT chatbot (Flask + Claude) with bundled knowledge base, CSRF protection, rate limiting, and encrypted session management
  • Deployed Federal Register and DOJ press release scrapers on DigitalOcean feeding into the live intelligence pipeline
Security Career (2014 — 2024)

Security & Reconnaissance Professional

US Army National Guard · State Corrections · G4S Security

  • Cavalry Scout (19D), Army National Guard — Enlisted at 17, trained at Fort Benning in reconnaissance, surveillance, and threat assessment in high-stress environments
  • Correctional Officer — Monitored and documented behavioral patterns in a controlled high-security environment, developing observational discipline and situational awareness
  • G4S Private Security — Physical security operations requiring strict protocols, environmental monitoring, and incident documentation
  • Core competencies across all roles: observational discipline, objective threat assessment, fact-based documentation, and separating actionable intelligence from noise

Work With Me

Available for contract work and full-time roles. Specialties include custom OSINT pipelines, government contracting automation (SAM.gov, state portals), document processing systems, AI-powered research tools, and Flask web applications with production-grade security.