I design agentic AI, retrieval, and evaluation workflows that connect LLM reasoning with reliable data, backend services, and clear engineering controls. My focus is not just generating answers—it is making AI systems easier to measure, debug, and improve.
A focused selection of research, nonprofit, and independent work across healthcare AI, enterprise retrieval, and multi-agent applications.
Contributing to a healthcare AI research initiative by building modular agents for knowledge retrieval, structured reasoning, and evaluation. The work emphasizes traceability, confidence-aware outputs, reproducible pipelines, and clear separation between evidence gathering and downstream decisions.
Developed a retrieval-augmented assistant over historical grant content using Azure OpenAI, Azure AI Search, hybrid retrieval, chunking, and metadata filters. The system was designed to improve access to funder-aligned context and reduce manual first-draft preparation.
Built a six-agent research workflow that separates planning, web discovery, content extraction, semantic retrieval, grounded writing, and evaluation. The architecture improves traceability and control by turning a single opaque generation step into a modular, inspectable system.
My work combines AI engineering with the practical disciplines that make systems dependable: data quality, modular architecture, retrieval evaluation, backend integration, and clear documentation.
Contributing to a healthcare AI research initiative focused on modular agent workflows for evidence-aware analysis. Building retrieval, structured reasoning, and evaluation components with an emphasis on provenance, confidence-aware outputs, and reproducible engineering.
Developed an Azure-based RAG assistant for grant workflows and created Python validation pipelines for operational data. Improved retrieval quality through chunking, metadata filtering, hybrid search, and iterative evaluation while reducing repetitive reporting effort.
Used Python, Excel, and Power BI to clean attendance data, automate recurring reporting, and improve visibility into participation trends across programs and operations.
I work across the full application layer—from retrieval and agent orchestration to APIs, evaluation, cloud services, and deployment-oriented workflows.
Multi-agent workflows, tool calling, stateful orchestration, structured outputs, prompt design, human-in-the-loop patterns, and failure-aware control flows.
Hybrid search, semantic retrieval, metadata filtering, chunking, re-ranking, vector databases, provenance, and retrieval-quality evaluation.
Grounding checks, relevance and faithfulness evaluation, LLM-as-Judge workflows, benchmark design, ambiguity handling, logging, and measurable iteration.
FastAPI, REST services, Pydantic, modular Python, async workflows, data validation, error handling, and production-oriented application structure.
PyTorch, scikit-learn, Hugging Face, Pandas, NumPy, feature engineering, classification, model evaluation, and data-quality workflows.
Azure OpenAI, Azure AI Search, AWS Bedrock, S3, Lambda, SageMaker, Docker, Kubernetes, GitHub Actions, CI/CD, and MLflow.
Graduate training in computer science supported by hands-on work across NLP, machine learning, retrieval systems, and backend engineering.
Graduate coursework included Applications of NLP, AI Ethics for Tech Leaders, and Research Methods for Human-Centered Computing.
Built a strong foundation in software engineering, machine learning, data science, algorithms, and applied artificial intelligence.
I am interested in roles where I can build reliable LLM applications, agentic workflows, retrieval systems, and production-oriented AI services—especially where evaluation and system quality matter.