Agentic Infrastructure
Revamping enterprise data lake build processes. Testing the boundaries of Claude Code to see how much boilerplate, scaling logic, and deployment architecture can be safely handed over to an autonomous agent.
Currently Working On
Evals over Syntax
Moving past traditional CI/CD. Because LLM outputs are probabilistic, deterministic testing is no longer enough. Writing robust, automated evaluation frameworks to ensure agentic workflows don't hallucinate destructive code into production environments.
Context Engineering
Building the pre-flight pipelines. An LLM is only as good as the context it holds. Exploring patterns to dynamically feed exact database schemas, historical logs, and architecture docs into Claude's context window before it begins executing a task.