Purity K.

Python developer & Data engineering | ETL Pipelines,Finance,LLM,Claude

I build production data systems that run correctly after I leave — pipelines, LLM workflows, backtesting engines, and Excel/VBA automation. Not prototypes. Not one-off scripts. Former Amazon Data Engineer. Doctor of Engineering, Data Engineering, University of Florida. WHAT I'VE DELIVERED IN PRODUCTION: → Claude API & LLM workflows — document automation and agent pipelines using the Anthropic API, integrated into existing data systems. → Weekly Tech Tracker — config-driven Python pipeline processing 280,930 rows across 66 companies every week. One config change adds a new grouping. No code deploy required. → Valuation Anchor Engine — 817-day backtesting system across AMZN, GOOGL, MSFT, ORCL. HAC/Newey-West corrected, FDR-adjusted. 100% in-sample hit rate on best-anchor signals — reported honestly as in-sample, not a forward guarantee. → Macro Data Pipeline — 20+ FRED series, zero manual refresh steps, live Looker Studio dashboard. Client scaled the engagement from 30 to 40 hours/week after the first delivery. → JST Service Awards — VBA Excel automation across 14 business units, 1,000+ employees. Full refresh under 60 seconds, zero formula errors, CHRO-approved. → National ML Infrastructure — Random Forest/XGBoost ensemble (0.89 AUC-ROC), deployed across 3 national projects, adopted by an international development partner as their primary screening tool. 47 regions, 160+ staff trained. WHERE I FIT BEST: - Claude & LLM integration — API workflows, document automation, RAG pipelines, agent workflows - Data engineering & ETL — Python, SQL, Snowflake, BigQuery, dbt, config-driven pipelines - Quant finance — backtesting, statistical validation (HAC/NW, FDR), financial modeling - Excel/VBA & BI automation — Power BI, Looker Studio, Tableau WHAT SEPARATES MY WORK: Config-driven by design — output updates when an input changes, no developer required after handover. QA runs automatically on every execution, not as an afterthought. I separate in-sample results from forward-tested ones and say so plainly, before a client finds out the hard way. Python · SQL · Anthropic Claude API · LLM Prompt Engineering · Snowflake · BigQuery · dbt · Google Apps Script · Looker Studio · Power BI · pandas/NumPy/scipy · statsmodels · scikit-learn · XGBoost · VBA Excel Doctor of Engineering, Data Engineering, University of Florida. Previously Amazon (data engineering, ETL, Power BI). Trilingual: English, Swahili, Spanish. Send me what you're working with — I'll give you a straight answer on fit, timeline, and cost before you spend a single connect.