I build data platforms, RAG systems and agents that people actually use. Five years in data and AI, most recently as the only data person in a 250-person procurement org at Stanley Black & Decker (2023 to 2026), where I worked directly with finance, engineering and suppliers.
Open to: Forward-Deployed Engineer, AI Engineer and Data Engineer roles in the US.
- A/B Test Readout Agent: CLI + Streamlit app that analyzes a two-variant experiment, runs a sample-ratio-mismatch check (SRM p < 0.001 forces "inconclusive"), and drafts a stakeholder readout with one GPT-5.6 call and self-checks it with a second. Python, MIT license.
- Customer Churn Prediction: Optuna-tuned XGBoost with SHAP explanations, served by FastAPI in Docker and deployed to AWS ECS/Fargate through GitHub Actions. 92%+ recall at under 10 ms inference.
- Causal inference on spaced-repetition data: live Streamlit app with a 61-day RCT (3,490 notes) and Double ML with 5-fold cross-fitting.
- Smart Collections prototype: a recommendation prototype for Adobe Lightroom built on synthetic catalog data.
- Reinforcement learning with Unity: earlier work.
Languages: Python, SQL, R, C/C++
Data: Snowflake, dbt, AWS (S3, Lambda, EC2, ECS/Fargate, QuickSight)
LLM: RAG, text-to-SQL and tool-using agents, LLM-as-judge evaluation, OpenAI and Anthropic models
ML and stats: scikit-learn, XGBoost, PyTorch, TensorFlow, MLflow, SHAP, A/B testing, causal inference (DoWhy, CausalForestDML, Double ML)
Tooling: Docker, GitHub Actions, FastAPI, Streamlit, Power BI
MS in Data Science and Analytics (Georgia State University, GPA 3.89) · BS in Computer Science (ABV-IIITM Gwalior) · Certifications: AWS Certified Cloud Practitioner, DataCamp AI Engineer, Coursera Machine Learning, DeepLearning.AI Deep Learning
⭐️ Exploring the boundary between causal inference, model interpretability, and real-world impact, one dataset at a time. Hit me up on any platform if you'd like to connect!