Building reliable evaluation for LLM assistants
Practical notes on relevance scoring, hallucination checks, and guardrail design for production-facing assistants.
Read morePhD Candidate (Expected May 2026) in Computer Science
AI Scientist Building Multimodal Systems for Vision, Language, and Real-World AI.
I build high-impact machine learning systems grounded in rigorous experimentation. My work spans LLM systems, object detection and tracking, media forensics, and edge-to-cloud deployment in real-world environments.
Computer Vision
Deep Learning
Media Forensics
ML Systems
LLM + RAG
Python · C++
Research snapshot
Publications & patents
4+ peer-reviewed · 2 patents
IEEE venues, applied vision systems, and medical AI innovations.
Toolbox
PyTorch · TensorFlow · Python · C++
RAG, reranking, agentic workflows, MLOps, and edge deployment.
Recent from site
Paper Reading List
- Loading Fetching latest papers...
Recent Blog Posts
- Loading Fetching latest posts...
At a Glance
A structured snapshot of my profile, current work, and the roles I am targeting.
Current role
Research Assistant, Office of Strategic Initiatives
Leading full-stack LLM assistant development with production-oriented RAG pipelines.
Primary focus
Vision, Forensics, and LLM Systems
From hypothesis-driven research to deployment-ready implementations.
Target roles
AI Scientist · Applied Scientist · Senior ML Engineer
Open to remote and relocation opportunities.
Academic timeline
PhD (Expected May 2026) Computer Science, AI-ML; MS Computer
Science; BE - Computer Science & Engineering
University at Albany, SUNY and SDMCET, Karnataka, India.
How I Contribute
I typically work in a three-part format to keep research quality high and delivery predictable.
01
Scientific rigor
I design experiments around clear hypotheses, baselines, and ablations so results are interpretable and actionable.
02
ML systems thinking
From data pipelines to model packaging, I build reproducible systems that move cleanly from notebooks to production.
03
Research communication
I turn experiments into clear reports, publication-quality visuals, and concise technical narratives for faster decisions.
Selected applied research
Flagship projects from healthcare AI, media forensics, and transportation analytics.
LLM Research Assistant Platform (GrantsMate)
Built a full-stack LLM assistant with hybrid retrieval, reranking, and agentic workflows for knowledge automation and decision support.
DARPA Semantic Forensics (SemaFor)
Developed semantic forensics pipelines for falsified media detection using object, text, and human-pose reasoning signals.
Railcar Detection, Tracking, and Yard Intelligence
Engineered a multi-camera rail analytics system with edge-control-center architecture using NVIDIA AGX Xavier for real-world operations.
Papers to Review
Current reading queue and paper topics I am actively reviewing for research and implementation.
Computer vision & forensics
- Scene text manipulation: robustness and synthetic data generation methods.
- Semantic forensics: cross-signal detection with text, object, and pose cues.
- Tracking and re-ID: methods for crowded and long-range industrial scenes.
LLM systems & evaluation
- RAG pipelines: retrieval quality, reranking, and grounded generation.
- Agentic reasoning: decomposition, memory, and tool orchestration quality.
- Safety evaluation: hallucination detection and response reliability metrics.
Blog Posts & Updates
Short technical updates from ongoing projects, experiments, and implementation lessons.
From computer vision experiments to deployable systems
A framework for moving from prototype notebooks to reproducible pipelines and team-ready deliverables.
Read moreVisitors around the globe
A live footprint of who’s stopping by and from which continent.
Dots scale as visits increase. Location is resolved on your device only.
Visits by continent
- North America—
- South America—
- Europe—
- Africa—
- Asia—
- Oceania—
Counts update instantly and stay anonymous—just a snapshot of global interest.
Let’s build something together
I thrive in teams that value rigorous experimentation and shipping. If you need an engineer who can translate research into dependable features, let’s talk.