Skip to content

Systems engineer, working where AI meets infrastructure.

I'm Alejandro Sierra, based in Austin. My work is the layer under AI products: running agents reliably, giving them budgets and boundaries, tracing what they do, and moving data to and from models without drama.

Sierralab is the name for what I build outside of work. Most of it is open source.

What I hold to

  1. Deterministic where it counts. Unbounded agent loops become runaway costs and drifting state. Orchestration needs finite transitions, explicit budgets, and a way back.
  2. Standards over glue. Ad-hoc bridges rot. Protocols like MCP for tools and A2A between agents keep systems inspectable after the demo.
  3. Trace before you trust. If an agent's behaviour isn't recorded, it isn't understood. Telemetry comes first, autonomy second.
  4. Local first, cloud when it matters. Fast feedback needs native tools and sandboxed local execution. Durable cloud infrastructure takes over for scale.

Tools

Languages
Rust, Go, TypeScript, Python
AI
AWS Bedrock, Google ADK, MCP, A2A, LanceDB, Ollama
Infrastructure
Lambda, Lightsail, S3, PostgreSQL, SQLite with Litestream
Interfaces
Next.js, Ratatui, GPUI

Outside work

A few times a year I take on an outside project: an organization with a real problem where AI and solid infrastructure could change the outcome. Write to hello@sierralab.net.