Selected work

Applied AI projects across healthcare, forensics, and industrial computer vision.

This portfolio reflects the projects on my resume, including DARPA-funded forensics, rail analytics systems, seizure detection research, and LLM-powered product systems.

Hire me for your team See resume

Project profile

Domains: healthcare AI, media forensics, rail transportation safety.
Methods: object detection, tracking, re-identification, LLM + RAG systems.
Delivery: reproducible research pipelines and deployable engineering artifacts.

Research portfolio

Flagship projects directly aligned with my resume and publication record.

LLM Assistant for Research Knowledge Automation

Designed a full-stack assistant for retrieval and Q&A with hybrid search, reranking, and multi-step agentic workflows for institutional research support.

LLM RAG Cross-Encoder Evaluation

See resume details GenAI

DARPA Semantic Forensics (SemaFor)

Developed semantic pipelines to detect falsified media using object-level cues, scene text, and human-pose evidence across manipulated content.

Forensics Detection Multi-signal Reasoning

See resume details Defense

Railcar Locomotive Yard Detection System

Built a rail-yard management platform using CV detection, tracking, and re-identification with edge-control-center architecture and NVIDIA AGX Xavier.

Edge AI Tracking Re-ID

See resume details Industrial AI

Seizure Detection in Animals (AMC-GE-UAlbany)

Developed system components for pose estimation, seizure-state prediction, and behavior analysis under induced and natural conditions.

Healthcare AI Pose Estimation Prediction

See resume details Biomedical

How I run research projects

The operating principles I use to move quickly while maintaining scientific quality.

Begin with baselines: establish reference performance before introducing complexity.
Evaluate carefully: track metrics, failure cases, and ablations in each cycle.
Document decisions: keep assumptions, choices, and results easy to audit.
Iterate to production: harden promising prototypes into maintainable systems.

Start a collaboration Read my notes