I build production-oriented AI and data systems with a focus on correctness, determinism, and deployment constraints.

My work emphasizes: • explicit system boundaries • failure-aware design • evaluation aligned with risk • reproducible data pipelines • infrastructure-aware ML deployment

LexBrief AI — Safety-First Legal NLP System
Hierarchical long-document NLP with deterministic decoding, clause extraction, rule-based risk engine, jurisdiction-aware logic, and production-safe fallback handling.

Python · Django · LegalT5 (PyTorch-based Transformer) · Rule-based systems · Docker · Evaluation design

Systemic Risk & Inequality Analytics Platform
Validation-first ingestion, immutable analytical artifacts, non-causal framing, regime-based synthesis, and leakage-resistant data workflows.

Python · Pandas · Statistical analysis · Reproducible pipelines · Data contracts

Languages: Python, Java
ML & Evaluation: NLP pipelines, metrics design, recall prioritization, latency profiling, failure analysis
Backend & Systems: Django, REST APIs, PostgreSQL, stateless service design
Data Engineering: Schema validation, dataset immutability, modular pipelines
MLOps: PyTest, Docker, CI-backed evaluation, YAML configs

• Production ML under deployment constraints
• Deterministic inference for regulated domains
• Guardrail-first AI system design
• High-trust analytical systems