Work
Selected projects.
Professional AI engineering and independent projects, including public code, tests, and implementation notes.
Professional practice · Environics Analytics
My AI engineering work covers access control for AI-connected search and tools, a tested remote agent-tool integration using OAuth2/PKCE, and evaluation workflows with human review. The authorization and evaluation work were independently owned; I contributed authenticated transport and reporting SDK work to a separate team-owned agent-tool platform.
Before moving into AI engineering in April 2026, I worked on recurring data-quality checks, release evaluation and anomaly investigations as a Research Associate (February 2024–April 2026).
Independent projects · public code
01AgentShield — evaluating injection attacks against AI agentsPaired prompt-injection experiments, tool authorization checks, and reproducible execution traces for LangGraph agents.2026
Public repository02Evaluating a RAG system by the ways it can failA reproducible document-QA experiment comparing a deterministic baseline with a bounded tool-using model.2026
Public repository03Testing incremental models with dbt and DuckDBA local DuckDB + dbt analytics pipeline that makes incremental correctness and data-quality failures visible.2026
Public repository04From expressive piano audio to editable notationAn audio-to-MIDI-to-score pipeline that makes timing preservation and notation loss explicit.2026
Public repository
Public repository02Evaluating a RAG system by the ways it can failA reproducible document-QA experiment comparing a deterministic baseline with a bounded tool-using model.2026
Public repository03Testing incremental models with dbt and DuckDBA local DuckDB + dbt analytics pipeline that makes incremental correctness and data-quality failures visible.2026
Public repository04From expressive piano audio to editable notationAn audio-to-MIDI-to-score pipeline that makes timing preservation and notation loss explicit.2026
Public repository