🎯 Founding Engineer @ AI Growth Labs
🧩 Previously Founding Engineer @ Aucctus
Builder of intelligent systems that reason, learn, and adapt.
Great AI systems aren't about making models smarter — they're about making the systems around them more deterministic.
LLMs are probabilistic engines. The craft is in architecting the constraints, context layers, and state machines that make their behavior predictable, traceable, and aligned with intent. A few things I've found to be true:
- Murphy's Law of LLMs: Anything a model can misinterpret, it eventually will — unless the system prevents it.
- Context before capability: Structured context improves reasoning more reliably than scaling model size.
- Determinism by design: Every agent should have an observable plan, bounded actions, and measurable outputs.
- Memory is architecture: How a system stores, retrieves, and expires knowledge is as important as the model itself.
My goal is to build reliable, explainable, and self-correcting intelligence — where uncertainty is contained by architecture, not hidden behind it.
Right now I'm most interested in agent memory architecture — specifically how systems store, expire, and retrieve context without ballooning cost or latency.
Beyond that:
- LLM & Agent Systems: ReAct orchestration, DAG-based workflows, GraphRAG
- Backend & Infra: FastAPI, Elixir/Phoenix, Redis, Celery, AWS, Docker
- Frontend: React, Next.js, LightningJS, WebGL-driven UIs for intelligent systems
- Tooling: OpenTelemetry, LangFuse, NL-to-SQL, multi-provider LLM pipelines
