Platform & AI Infrastructure Engineer building reliable distributed systems, developer infrastructure, and AI platforms in Go, Rust, Python, and TypeScript.
I focus on the parts of software that have to remain understandable when systems fail: durable execution, data integrity, privacy boundaries, measurable search, and operational control planes.
Sluice — LLM infrastructure
An OpenAI-compatible gateway for controlling cost, privacy, and provider failure. Its evidence includes reversible PII redaction, semantic caching, budget policy, provider failover, and audit logs that can be replayed.
Skald — durable execution
A workflow engine that lets ordinary Go functions survive crashes, deploys, and long waits. Append-only histories, deterministic replay, recovery simulations, timers, workers, and operational tooling make its guarantees inspectable.
Proxima — vector search
A vector search engine built from first principles in Rust, including HNSW, IVF-PQ, quantization, filtered search, and write-ahead-log recovery. Recall is measured against exact search rather than treated as a marketing claim.
Basalt — analytical storage
Readable columnar database components in Rust: arrays, compression codecs, segment storage, a SQL parser, binding, and logical planning. Its README states the unfinished executor and optimizer explicitly so implemented scope is clear.
Nabla — machine-learning foundations
A compact deep-learning framework built on NumPy with reverse-mode autodiff, layers, optimizers, and training loops. Every differentiable operation is checked against numerical gradients, including broadcast reduction and shape contracts.
Meridian Control Plane — governed operations
A TypeScript control plane combining a React console, multi-tenant API, durable worker, approval gates, audit events, and outbox-shaped delivery. It demonstrates how backend guarantees and an operational interface fit together.
| Area | What I build | Primary tools |
|---|---|---|
| Distributed systems | Durable workflows, retries, idempotency, recovery | Go, PostgreSQL, SQLite |
| AI infrastructure | Model gateways, privacy controls, search, evaluation | Go, Rust, Python |
| Data systems | Storage formats, indexing, planning, quality gates | Rust, Python, SQL |
| Product infrastructure | Multi-tenant APIs, workers, audit trails, consoles | TypeScript, Fastify, React |
- Start with explicit invariants, failure modes, and honest scope boundaries.
- Keep critical mechanisms observable through tests, benchmarks, audit events, deterministic fixtures, or replayable histories.
- Prefer small dependency surfaces when implementing a subsystem is the point; use established libraries when the product problem is elsewhere.
- Treat documentation, security policy, reproducible setup, and CI as part of the system rather than finishing touches.
Resumen en español
Soy ingeniero de plataformas e infraestructura para IA. Construyo sistemas distribuidos, motores y herramientas que hacen explícitas la durabilidad, la privacidad, la integridad de datos y la operación bajo fallos. Mi trabajo combina Go, Rust, Python y TypeScript, con pruebas reproducibles, documentación técnica y límites honestos.