Research
Context-Efficient RAG for Multi-Hop Reasoning
Reasoning-aware retrieval compared with conventional chunks across evidence coverage, context efficiency, retrieval quality, and answer quality.
Data Scientist · Applied AI
Turning messy data into decision-ready systems.
Explore my work
Research, systems, and applied case studies.
Each piece opens into a full article covering the question, methodology, architecture, evaluation, results, tradeoffs, and source code where public.
Research
Nori Benchmark
Controlled multi-seed evaluation of a tabular foundation model, including context-size sweeps and irrelevant-feature stress testing.
AI Systems
Enterprise AI Workload Intelligence
Workload-aware model routing across cost, quality, reliability, latency, sparse history, and confidence-aware decisioning.
Applied ML
Legal Charge Classification
A multi-stage system combining semantic retrieval, supervised ensembles, attribute re-ranking, calibrated confidence, and targeted LLM adjudication.
About me
I like problems where the data is imperfect and the decision still has to be made.
I’m a Data Scientist based in New York. My work has spanned public-sector data systems, retrieval and RAG, AI quality, causal inference, forecasting, and model evaluation.
I tend to build around the full decision loop: data → evidence → model → confidence → action → evaluation.
More about me ↗Experience
Applied data science across public systems, AI quality, and decision intelligence.
CUNY Institute for State & Local Governance
New York, New York, United States
AI classification systems, demographic data infrastructure, and public-sector research.
NYC Office of the Mayor
New York, New York, United States
Procurement analytics, RAG systems, risk modeling, and forecasting.
Baruch College, CUNY
New York, New York, United States
Analytics instruction across forecasting, regression, SPC, and Lean Six Sigma.
Gentrainer, Meyers Workforce Solutions, LLC.
New York, New York, United States
AI training-data quality, model evaluation, feature engineering, and MLflow workflows.
Think & Learn Private Limited (BYJU'S)
Bengaluru, Karnataka, India
Causal inference, uplift modeling, personalization, and decision intelligence.