Pereira, Colombia · UTC−5 · Open to remote and contract work
Backend systems,
AI agents, and
one robot arm.
I’m Victor Cano, a self-taught solutions architect. I’ve been writing code since I was thirteen and getting paid for it since I was sixteen — mostly Go, Rust, Python and TypeScript. Everything I’ve had running is below, drawn the way I’d draw a distributed trace: overlapping, because it always is.
- Teaching myself to shipStarted at 13. PHP first, then everything else. No degree, no bootcamp.
- First paid workOdd jobs at 16 — small sites and fixes, while still learning in the open.
- Freelance engineeringContract backend and mobile work, mostly LatAm and US clients.
- IA PortafolioCo-founded a design-and-technology studio. Three partners.
- Círculo Militar PerúMembership and reservations platform. PHP / Laravel, PostgreSQL.
- Brooklyn InfrastructureInfrastructure tooling for Tango Software. Go services, AWS.
- RSM-NetRecursive submatrix memory network for continual learning.
- Pisasuave · Eco · LumeThree small SaaS bets: DM automation, meeting capture, guest recognition.
- The Magic AgentsMulti-tenant WhatsApp AI agents. Go, gRPC, pgvector, RLS.Read the case study →
- FaroSelf-hosted observability. Rust / Axum / ClickHouse / SvelteKit.Read the case study →
- CodrexRust coding-agent runtime. Cheap worker model, strong orchestrator.Read the case study →
- arbscanGo scanner for cross-book arbitrage across Polymarket and Cloudbet.
- MANO6-DOF arm on a mecanum base. Raspberry Pi 5, on-camera inference.Read the case study →
- Research
- Client work
- Own product
- Hardware
Selected work
A multi-tenant platform that lets a business run an AI agent on WhatsApp without handing its customer data to anyone else.
Self-hosted observability — traces, logs and metrics on ClickHouse — built because hosted APM is priced for companies with a finance department.
A Rust coding-agent runtime that pairs a cheap worker model with a strong orchestrator, to find out how much of agentic coding is really about the model.
A 6-DOF arm on a mecanum base that takes instructions in plain language — Raspberry Pi 5, on-camera inference, and a spatial-reasoning model in the loop.
How I work
- I read the source
- When a library misbehaves I go into its code rather than around it. Most of the hard problems I have solved were an afternoon spent inside someone else’s repository.
- Multi-tenancy is a day-one decision
- Row-level security in Postgres, a tenant id on every span and every log line. Retrofitting isolation into a live product is the most expensive rewrite I know of.
- Instrument, then optimise
- I built my own observability platform because I got tired of guessing. If I cannot see where the time goes, I do not have an opinion about where the time goes.