AI systems,
stress-tested.
Research-driven work exploring safety, reliability, benchmarking, and behavior under stress in large-scale software and AI systems.
Papers & investigations.
AgentLens
Open-source developer tool that diagnoses AI-agent failures — classifying root causes (tool selection, loops, cascades, context pollution, state drift, overflow) and computing the wasted-token cost of each. Ships Python & TypeScript SDKs, a FastAPI server, and a CLI; integrates with Anthropic, OpenAI, LangGraph, CrewAI, AutoGen, and PydanticAI via runtime instrumentation.
LLM-CODEGEN
Advanced code generation system using Large Language Models.
Large-Scale Bug Prediction
Benchmarking bug prediction across 700k Python projects.
AI Recommender
Intelligent recommendation engine for personalized content.
AI in Software Engineering
Research on applying AI techniques to software engineering problems.
Research Paper
Academic research paper repository.
VLM Failure Modes
Analysis of failure modes in Vision-Language Models.
VLM Adversarial Defense
Defense mechanisms against adversarial attacks on VLMs.
Amnesic VLM Defense
Amnesic defense techniques for Vision-Language Models.
AI Impact on Jobs
Analysis of AI's impact on the job market.
Sentiment Steering GPT
Steering GPT output sentiment.
Research lives on GitHub.
Papers, benchmarks, and experiment code are versioned alongside the implementations.