I build at the intersection of AI/ML and production software — from LLM reliability research to full-stack AI platforms. Currently working on codebase intelligence tooling and early hallucination detection in language models.
|
🧠 DevLens AI Codebase intelligence from zero-cost infra AI-native platform that parses, graphs, and understands entire codebases — Tree-sitter for structure, pgvector for semantic search, LLMs for chat & architecture views.
|
🔬 Early Hallucination Detection Catching LLM lies mid-sentence Multi-signal fusion system combining hidden-state probes, attention patterns, logit uncertainty, and semantic entropy to flag unreliable generations before they finish.
|
|
🤖 AI Job Application System Automating the job hunt with LLMs End-to-end pipeline that scrapes listings, evaluates relevance, and generates tailored ATS-optimized resumes and cover letters.
|
🕵️ Deepfake Detection System Catching what the eye can't Real-time image & video deepfake classification with a preprocessing pipeline, web inference, and a Chrome extension for browser-level detection.
|