Regional sky model observer context
Human-centered AI, engineered for the real world
Abhineet Pandey
Ph.D. candidate and AI/ML researcher building LLM systems, agentic workflows, and applied computer vision
I build dependable AI systems across LLM-powered workflows, retrieval-augmented generation, structured evaluation, multimodal AI, computer vision, media forensics, and edge deployment.
Best fit for teams working on multimodal AI, computer vision, trustworthy LLM applications, applied research, and ML platform delivery.
AI ScientistApplied ScientistSenior ML Engineer
Professional evidence
Download resume Role fit, delivery evidence, and research depth
Role fit, delivery evidence, current availability, and contact paths are organized for a concise review.
Architecture decisions, trade-offs, system boundaries, and production constraints are foregrounded.
Publications, evaluation practice, scientific framing, and domain collaborations are foregrounded.
Applied AI capability map
Case studies
Review full project archive Representative systems and applied AI work
GrantsMate
ProblemFunding discovery and research-admin workflows are fragmented and slow.
BuiltEnd-to-end AI platform for opportunity matching, collaborator discovery, and knowledge workflows.
ImpactLive startup platform with CTO-level product and engineering ownership.
Filippo
ProblemPatent comparison requires reading dense technical documents with traceable evidence.
BuiltAI-powered patent analysis and comparison product.
ImpactLive public product built end to end as sole developer.
LLM Research Assistant Platform
ProblemResearch teams need grounded answers across large institutional knowledge collections.
BuiltHybrid retrieval, reranking, agentic workflow, and evaluation components.
ImpactProduction-oriented assistant architecture for dependable knowledge access.
DARPA Semantic Forensics (SemaFor)
ProblemFalsified media detection needs semantic evidence beyond low-level artifacts.
BuiltComputer vision pipelines combining object, scene-text, and human-pose evidence.
ImpactResearch contributions in high-stakes semantic media forensics.
Trajectory
Professional trajectory
Summer Graduate Fellow, Agentic AI and Red Teaming
Designing secure local-compute agentic AI workflows for proposal red-teaming, structured feedback, and responsible institutional evaluation.
CTO and sole developer, GrantsMate
Building AI product workflows for grant discovery, collaborator matching, and research administration.
Sole developer, Filippo
Shipping an AI-powered patent analysis and comparison product.
LLM Research Assistant Platform
Hybrid retrieval, reranking, agent workflows, and evaluation for grounded research automation.
Healthcare computer vision research
Pose estimation and behavior analysis workflows for seizure-state detection.
DARPA Semantic Forensics
Semantic media-forensics pipelines using object, text, and pose evidence.
Rail-yard intelligence
Detection, tracking, re-identification, and edge deployment for industrial vision.
Operating principles
Consistent strengths across research and engineering
“Bridges research thinking and product execution without losing rigor.”
Research and product execution“Strong fit for teams that need experiments turned into maintainable systems.”
Applied AI delivery“Comfortable across vision, retrieval, applied ML, and full-stack delivery.”
Technical breadth Products & applied AI
Review project portfolio Building the product, the platform, and the intelligence layer
GrantsMate
An AI platform for grant discovery, collaborator matching, and research-administration knowledge workflows.
Read case study →Filippo
An AI-powered patent analysis and comparison application.
Read case study →LLM Research Assistant Platform
A full-stack assistant for retrieval, question answering, and research knowledge automation using hybrid search, reranking, and agentic workflows.
Read case study →DARPA Semantic Forensics (SemaFor)
Semantic forensics pipelines for detecting falsified media using object, scene-text, and human-pose evidence.
Read case study → Published developer tools
Explore all software Software shipped to the VS Code ecosystem
AI developer tooling
Agent Inspector
Monitor AI coding agent context, visible files, token usage, and trust signals in VS Code-compatible editors.
Remote compute
Pipe Explorer
Manage remote compute and service pipes from VS Code, including Slurm Jupyter, TensorFlow streams, and custom services.
Research tooling
TexPilot
Professional LaTeX workspace with integrated compilation, logs, preview, and local-first AI assistance.
Academic productivity
TeXPilot LaTeX Workspace
A VS Code-compatible LaTeX workspace for students and researchers with local-first AI and paper-writing workflows.
Research record ForensiText: ROI-Guided Multimodal Forensics for Scene Text Tampering Detection and Localization 22nd International Conference On Advanced Visual And Signal-Based Systems (AVSS), 2026 Integrating manual preprocessing with automated feature extraction for improved rodent seizure classification Epilepsy and Behavior, 2025 TextSleuth: A New Dataset and Baseline for Scene Text Manipulation Detection 2024 IEEE 7th International Conference on Multimedia Information Processing and Retrieval (MIPR), 2024 Review publications and patents →
Representative publications
Toolbox
Research depth, production instincts
LanguagesPython / C / C++ / Java / CUDA / MATLAB / SQL
MLLarge Language Models / Computer Vision / Deep Learning / Multimodal AI / Model Evaluation / Hallucination Analysis
LLM systemsRAG / Agentic Workflows / Prompt Engineering / Structured Outputs / Hybrid Retrieval / Cross-encoder Reranking
PlatformsDocker / Linux Servers / GPU Servers / PostgreSQL / Neo4j / NVIDIA Jetson / Ollama
Live visitors
Loading visitor data…
Research to deployed systems
Email Abhineet AI systems for teams that need rigor, reliability, and delivery
I translate research ideas into reproducible experiments, maintainable software, and deployment-ready AI systems.