Skip to content
View JohnnyWilson16's full-sized avatar

Block or report JohnnyWilson16

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
JohnnyWilson16/README.md

Hi, I'm Johnny Wilson Dougherty

AI/ML engineer building agentic RAG systems, LLM-assisted data pipelines, and data-quality tooling.

LinkedIn Email


πŸ“Œ About Me

  • πŸ›‘οΈ Maintainer of FreshData: explainable data cleaning for pandas and Polars, published on PyPI as freshdata-cleaner. I wrote most of the codebase (~480 commits, 95 merged PRs) and review outside contributions.
  • 🌐 13 merged upstream PRs in projects I don't maintain, including CPython, pandas, TensorFlow, Ultralytics, Docling and AutoGPT. Full list below.
  • πŸ› οΈ Python, FastAPI, Pydantic, pandas, Polars, DuckDB, LangChain, CrewAI, ChromaDB, Ollama, Pytest.
  • 🎯 Open to full-time AI/ML roles (remote or onsite) and freelance GenAI / data engineering work.

πŸš€ Projects

πŸ›‘οΈ FreshData (maintainer)

  • Explainable, automated data cleaning for pandas and Polars, with DuckDB support.
  • Schema-preserving Polars/DuckDB bridges, semantic and per-field validation, text sanitizers and privacy filters.
  • Benchmark suite, streaming telemetry, and PyPI Trusted Publishing CI/CD.
  • Looking for contributors: see the open issues.
  • Agentic RAG over financial reports (CrewAI, LangChain, ChromaDB) that plans queries, checks evidence and cites sources.
  • Local-first: inference runs through Ollama, and the config rejects non-local LLM endpoints.
  • Three-tier guardrails (input, retrieval confidence, output grounding) with a pytest suite.
  • Detects upstream schema drift and remaps column headers at runtime instead of failing the job.
  • Layered resolution: exact match β†’ mapping cache β†’ heuristics/fuzzy match β†’ LLM fallback.
  • LLM prompts contain only column names and expected types, never row values.
  • Customer complaint triage with supervised NLP classifiers, served through FastAPI.
  • Automated routing, sentiment breakdown and reporting.

🌐 Merged Upstream Contributions

Project PR What it fixed
CPython #158962 asyncio: the threaded child watcher no longer blocks the event loop on macOS when a child process is stopped (SIGSTOP) rather than exited (gh-158287).
pandas #70838 Python CSV engine: custom converters that return unhashable values (lists, dicts) no longer crash type inference (#56848).
TensorFlow #129229 tf.data: from_tensor_slices / from_tensors convert host arrays on CPU instead of allocating GPU memory during dataset construction (#71744).
Ultralytics #26711 Clear error instead of a TensorRT assertion when exporting INT8 engines with nms=True on JetPack 6 (#26655).
Docling #4240 Safe fallback when Pydantic's serialize_as_any reports a false circular reference, which was crashing conversions (#4229).
AutoGPT #14875 Copilot web_fetch: keeps main page content instead of truncating raw HTML early, moves HTML cleaning off the event loop, and warns on JS-only page shells (#14828).
Headroom #3579 Proxy budget limits now enforced on OpenAI and Gemini routes and on WebSocket frames (#3374).
SkillSpector (NVIDIA) #903 Stops flagging HTTP DELETE verbs in REST route tables as destructive filesystem deletes (#832).
Semantica #1603 Nested metadata filtering in the MCP handle_query_decisions handler (#1247).
Tel-Agent #243, #244, #248, #252 Mattermost channel transport, gateway auth fix, mention cleanup and locale updates.

All merged PRs: github.com/pulls?q=author:JohnnyWilson16+is:merged


πŸ“¬ Contact

Pinned Loading

  1. agentic-rag-system agentic-rag-system Public

    Production-ready End-to-End Agentic RAG System using LangChain, ChromaDB, and CrewAI

    Python

  2. indian-pharmacy-analytics indian-pharmacy-analytics Public

    SQL-first analytics on 253,973 Indian medicines β€” stock monitoring, expiry risk, seasonal demand, and supplier optimization. SQLite + pandas + Faker.

    Python

  3. self-healing-data-pipeline self-healing-data-pipeline Public

    Autonomous self-healing data pipeline framework using schema caching, heuristics, and LLM semantic reasoning

    Python

  4. FreshCode-Org/freshdata FreshCode-Org/freshdata Public

    Automated, explainable data cleaning for pandas and Polars, repair messy tabular data safely and see exactly what changed.

    Python 6 19