$ whoami
Senior Full Stack & AI EngineerAgentic Systems, Backend & APIs
I build software where the business rules are as hard as the code: multi-tenant SaaS, serverless backends on AWS, RAG features and multi-agent AI systems with human approval gates. 5+ years taking products from data model to production.
$ cat engineering-principles.md
How I build
Patterns that keep showing up in the systems I build. Lessons from the work, not slogans.
Define the problem first
Projects fail on undefined problems, not on code. Scope and architecture come before implementation.
Isolate by default
Per-tenant databases and request-time injection shrink the blast radius and keep tenant logic out of business code.
Keep the request path short
Inference, document processing and heavy jobs run off the request path, so latency holds under load.
Assume upstream will fail
Every third-party service gets fallbacks and retries. A provider outage should never become a product outage.
Budget AI like any backend
LLM features ship with cost, latency, evaluation and failure budgets, like any other service.
Automate the path to prod
Separate dev, test and production environments, wired together with CI/CD, keep releases boring and predictable.