01 / What I bring
Production AI, and the discipline to keep it honest.
Each of these is something I have shipped and operated, not something I have read about. The line under each is the concrete evidence.
Retrieval & knowledge infrastructure
RAG over real corpora: embeddings, vector search, deduplication, chunk scoring, context and memory design, multi-source ingestion with quality gates.
102k vectors · 13k documents · 8 sourcesAgentic systems & orchestration
Multi-agent workflows with approval layers, tool dispatch and audit logging — built so a wrong answer is caught before it becomes a wrong action.
12+ concurrent agents · 200+ jobs → one schedulerData & automation engineering
Ingestion pipelines, API and CRM integration, workflow automation, Discord and Telegram bots, email and social automation.
20+ source adapters behind one contractInfrastructure & operations
VPS administration, containers, service supervision, VPN, GPU inference, monitoring and alerting. I deploy and run what I build.
Linux + Windows · CUDA · Tailscale · nginxData stores
Schema design and operational use across relational, document, cache and vector stores — chosen for the access pattern, not for novelty.
PostgreSQL · MongoDB · Redis · Qdrant · DuckDBEvaluation & quality gates
Benchmark design, regression gates and measurement discipline — the difference between "it looked right" and "it was verified".
Fine-tuning experiments · eval frameworks02 / The record
Dated honestly, including the parts that are short.
I have been building AI systems full-time for about fourteen months. Before that I spent four and a half years as a fundamental analyst in crypto markets. I would rather you read that accurately from me than infer something generous and discover otherwise.
03 / Verify it
Falsifiable, or it didn't happen.
Everything below is something you can open in another tab right now. I would rather be checked than believed.
- Live product vocax.ai — a multi-tenant SaaS I designed, built and deployed. Live in beta and taking payments.
- Open source autoresearch-orchestrator-pattern — a clean-room agentic research-loop orchestrator.
- Open source discord-voice-transcribe — voice capture and transcription for Discord.
- Open source vimeo-mcp-lite — a small, focused MCP server.
- Writing The article archive — written and published by a pipeline I built, which also gates its own quality. The system that publishes it is the same kind of system I am describing above.
- This site tristanhazelwood.com — the longer, technical account of the same work, including the parts that failed.
04 / Fit
Where I am useful, and where I am not.
I am strongest where a system has to run unattended and stay correct — retrieval pipelines, agent workflows, automation with real consequences, and the observability that tells you when any of it drifts. I work end to end: I will design it, build it, deploy it, and be the one who gets paged.
That range covers a lot of ground: greenfield build, hardening something that already exists, evaluation and research work that needs a rigorous hand, or getting a vendor platform genuinely productive rather than merely purchased. I learn systems quickly and in public — I taught myself distributed systems to a professional standard, then did the same with AI.
The one real constraint is location. I am based in Coffs Harbour NSW and work remotely across Australian and international hours. For a hybrid or on-site role I am open to it, but a relocation allowance or travel expenses would need to be part of the conversation.
Open to permanent employment or contract, whichever suits the team.
05 / See it
Or watch me teach it.
A six-minute cut from an AI literacy course I built and delivered for a client — a non-technical audience, live sessions, no script. If you would rather see how someone explains a thing than read their claim that they can, this is that.
The full sessions, unlisted, if you want the long version:
- 16:13 AI Foundations: LLMs, Automation & Agents — the opening session: what a language model actually is, and where automation ends.
- 14:35 AI Agents vs Raw LLMs: Beyond the LLM — where a raw model stops and an agent begins.
- 30:57 Choosing the Right LLM: A Model Selection Framework — picking a model by task rather than by hype.