John de Graft-Johnson

Experience —

Eight years across data engineering, applied ML, and AI platform delivery — recently focused on evaluation frameworks, fine-tuning, and responsible-AI tooling.

1. 01

   AI Engineering Lead · Swain Solutions LLC

   2025 – 2026

   Washington, D.C.

   • › Built LLM evaluation harnesses end-to-end in Python — deterministic judges, paired LLM auditors, weighted composite scorecards — covering fairness, factuality, and compliance dimensions across regulated proposal and clinical pipelines.
   • › Fine-tuned open-weights models (Gemma family) with LoRA and preference-optimization workflows; instrumented training with red-teaming harnesses (inspect_petri), Fairlearn fairness audits, and LangSmith observability.
   • › Owned the medallion data architecture (dbt-duckdb, bronze → silver → gold) and GPU inference path (CUDA, sub-1ms latency) for a nine-service Azure-native AI platform, including federated learning across 217 tickers and a live calibration model over 19K+ rows.
   • › Embedded a thirteen-point responsible-AI framework (HIPAA, NIST, CMMC, MHRA GMLP) into CI gates; every deployment ships with an evidence pack rather than a checklist.
   PythonLLM evalLoRA / SFTRed-teamingdbtAzure
2. 02

   Product Analytics Manager · W.R. Grace

   2024 – 2025

   Columbia, MD

   • › Designed evaluation and drift-detection frameworks for enterprise forecasting models; reduced forecast-to-actual variance by 18% through systematic empirical iteration rather than ad-hoc tuning.
   • › Owned the medallion data architecture (bronze / silver / gold) and feature engineering layer; coached analysts on model evaluation, feature selection, and production deployment standards.
   • › Translated ambiguous business questions into reproducible Python experiments and decision-ready artefacts for Finance, Operations, and the executive team.
   PythonModel evalDrift detectiondbtForecasting
3. 03

   Senior Data Analyst · W.R. Grace

   2022 – 2024

   New Orleans, LA

   • › Directed an enterprise revenue-intelligence program covering $600MM+ in monthly performance; built statistical forecasting models that reduced planning cycle time by 25%.
   • › Engineered production data pipelines (Python, SQL, dbt, Airflow) with data-quality gates, partitioning strategy, and CI/CD for analytics infrastructure.
   • › Applied experimentation frameworks for customer segmentation and cohort analysis; used model-driven evidence to influence go-to-market prioritization.
   PythonSQLAirflowForecastingExperimentation
4. 04

   Data Engineer · W.R. Grace

   2019 – 2022

   Baltimore, MD

   • › Established the data-integrity, governance, and compliance programs later scaled across the enterprise; built audit frameworks for regulated reporting environments.
   • › Designed Power BI dashboards with automated anomaly detection; data-driven interventions cut system downtime by 28% and lifted SLA compliance.
   • › Raised data accuracy by 14% through predictive modelling, outlier remediation, and systematic root-cause analysis.
   PythonAnomaly detectionGovernanceSQLPower BI

Selected Projects ★

Each project is shipped end-to-end — data pipeline, model, governance, deployed UI. Cards with an architecture map can be expanded inline.