I build Python tools for data analysis and data quality, and contribute to open-source agent integrations.
Contributed the AutoGen adapter for AgentFlow/HiveMind, supporting single-agent and team execution, streaming, and tool events. The integration includes documentation and tests.
Merged PR #80 · Upstream implementation · Unit tests · Integration guide
Python · AutoGen · Agent integration · Testing
A tool for investigating tutorial drop-offs, cohort churn, and duplicate event tracking. It combines SQL queries and cohort comparisons into reports with traceable evidence, using synthetic player data.
Demo GIF · Example report · Architecture · Query controls
Python · SQL · Pandas · SQLite · Streamlit · Plotly · MCP
Automated checks · 40-case offline evaluation of the tools and replay workflow.
A case study of how FGO JP events relate to public attention over time, with time-based model evaluation and anomaly review. The analysis uses public attention proxies; player activity and revenue are outside its scope.
Case study · 29-page report · Reproducible core · Evidence boundary
Python · Pandas · NumPy · SciPy · Time series · Data product
Evaluation: 2022–2023 training data · 2024 model selection · 2025 held-out test · Automated checks
A command-line tool that checks JSONL event data for missing fields, duplicate IDs, invalid timestamps, and business-rule violations. It uses the Python standard library and produces reports and exit codes for use in CI.
Contract example · Validation engine · Regression tests · Roadmap
Python · Data contracts · Data quality · JSONL · CLI · GitHub Actions
Automated checks cover valid and invalid event fixtures. Reports are available as JSON, Markdown, text, or JUnit XML.
A browser editor for turning text and images into EPUB 3 books. Includes chapter ordering, covers, and project save/restore; files are processed locally.
EPUB/ZIP core · Validation tests · Project README
JavaScript · HTML · CSS · EPUB 3