Open to opportunities

GlitzLitzenberg

AI/ML Engineer · End-to-end product shipping: zero-to-production TypeScript/React/Node monorepos with CI/CD · LLM/AI integration & cost governance: Claude API, conversation memory, spend metering, prompt architecture

St. Petersburg, FL, US
Book intro with Glitz
AI/ML EngineerAI Product EngineerConsultantDeveloperEngineer+4 more
01 — Positioning

Where Glitz is headed.

Self-taught full-stack engineer who designs and ships complete products from zero to production, combining frontend/backend expertise with deep ML fundamentals (transformer architecture, LLM training). Built a 15.5K-line conversational AI platform live in production and trained a GPT-style language model from scratch—bridging product velocity with machine learning depth.

What Glitz brings
  1. 01End-to-end product shipping: zero-to-production TypeScript/React/Node monorepos with CI/CD
  2. 02LLM/AI integration & cost governance: Claude API, conversation memory, spend metering, prompt architecture
  3. 03Machine learning fundamentals: transformer architecture, tokenization, training loops, PyTorch from scratch
  4. 04Real-time systems: WebSockets, Socket.io, WebRTC signaling, multi-user synchronization
  5. 05Full-stack security: JWT/bcrypt auth, RBAC, rate limiting, server-authoritative design patterns
02 — Proof

The receipts.

Numbers first. Every metric below is something Glitz shipped or drove — not a claim, an outcome.

  • Proof 01
    Next.js 16/React 19/Prisma 7 MVP with KYC, Stripe-ready payment abstraction, buyer/creator roles
    CreatorVault escrow marketplace
  • Proof 02
    8.5M-token corpus from 10 heterogeneous sources with robots.txt enforcement, per-domain rate limiting, resumable crawl state
    Data pipeline & web scraping
  • Proof 03
    15,500
    LOC TypeScript monorepo, zero to production in 6 weeks, live in production with Claude API cost governance
    Echoes of History conversational AI platform
  • Proof 04
    Decoder-only transformer from scratch, 8.5M-token corpus, validation loss reduced from 9.7 to 4.0
    GPT-style language model training
  • Proof 05Open ↗
    Live demo deployed on Vercel
    GuilDev governance platform
03 — Role fit

Operating range.

Where Glitz plugs in fastest — sorted from strongest fit to adjacent. Not a wishlist; a routing map.

  • primary fit
    AI/ML Engineer
    Mid
  • primary fit
    Full-Stack Engineer
    Mid to Senior
  • secondary fit
    Product Engineer
    Mid to Senior
  • adjacent fit
    Backend Engineer
    Mid