Mike Hall

Architected and operating sovereign production platforms across Ruby on Rails 8, PostgreSQL 18 (pgvector), TypeScript, and local Metal GPU AI inference runtimes, combining deterministic state engines, autonomous media processing pipelines, and distributed OpenTelemetry observability.

• Autonomous Media Pipeline & Local AI RAG Microservice: Architected a Rails 8, Solid Queue, and PostgreSQL 18 (pgvector) media ingestion and publishing platform with on-device Metal GPU Whisper transcription, hybrid lexical and vector semantic search via Reciprocal Rank Fusion (RRF), local LLaMA 3.1 RAG synthesis via RubyLLM, OpenTelemetry distributed tracing, Ahoy analytics, and PgHero query tuning across 270+ processed video masters.
• Forensic Audio DSP Restoration & Sovereign Verification (dehum): Engineered a deterministic 5-stage FFmpeg audio restoration and clock-drift synchronization pipeline (atempo, adeclick, highpass, afftdn, and 8-harmonic 60 Hz notch cascade) eliminating electromagnetic interference and corona discharge crackle recorded under 345,000-volt transmission lines while preserving lossless video streams.
• Technical Conversation Archive & Knowledge Graph: Built and curated a contract-validated oral history platform preserving 218 technical interviews with Software Craftsmanship, Ruby, and JVM practitioners, enforcing build-time schema verification, speaker diarization, and semantic knowledge graph cross-linking across 500+ indexed pages.
• Deterministic State Engine & Verification Platform (Phalanx Duel): Engineered a server-authoritative real-time tactical game platform in TypeScript, Node.js, PostgreSQL, and WebSockets featuring a pure deterministic rules engine, append-only replayable action ledger, OpenTelemetry trace propagation, and property-based CI verification gates across the complete state lifecycle.
• Multi-Source Ingestion & Prompt-Injection Defense (WWWorkRemote): Designed a Rails 8 and pgvector semantic matching platform with a Manifest V3 Chrome extension across 16 ATS adapters and a 4-stage prompt-injection guardrails pipeline screening untrusted scraped text before local LLM evaluation.

Skills: Ruby on Rails, PostgreSQL, pgvector, Solid Queue, RubyLLM, Local-First AI Inference, Whisper & Speech-to-Text, FFmpeg & Audio DSP, OpenTelemetry, TypeScript, Node.js, WebSockets, Deterministic Simulation, Prompt Injection Defense, Chrome Extensions (MV3), CI/CD