Skip to main content

Projects

Desktop software, developer tools, and experiments in computer vision.

HelpMe generates and validates a terminal plan for opening Brave.
2024–2026Demo

Architected and coded HelpMe, a terminal-based CLI tool that uses advanced AI Agents to automatically generate, validate, and execute step-by-step Linux commands from user instructions. Reduced manual effort by 75% for complex Linux tasks.

PythonLinuxLLM APIs
SnapSort’s desktop interface displaying a local photo library.
2025–2026Demo

Offline AI desktop app for face tagging and semantic image search, processing 2.5–3 images/sec on a 4-core CPU with YuNet, SFace, CLIP, and FAISS.

ElectronReactTypeScriptPython
Kova workspace showing topic highlights beside the Chrome side panel with a 94% research relevance score and scoped RAG answer.
2026

Intelligent web highlighter that scores page relevance before you read, isolating captured insights into strictly scoped AI workspaces with topic-isolated RAG chat.

Next.jsTypeScriptFastAPIPostgreSQL
Illustrative SOMU chat showing a sample support conversation.
2024

Constructed a robust system combining live human interaction and an AI-powered chatbot for immediate assistance in stress management; optimized service delivery efficiency, resulting in positive feedback from over 200 users.

NextJSTailwindShadcnPrismaPostgreSQLLiveKit APIClerkGemini API
Flow diagram showing a goal, API retrieval, and an ordered action plan.
2023–2024

Implemented the Reverse Chain algorithm to enhance LLM capabilities in multi-API action plan, integrating langchain, FAISS (vector database), BERT embeddings, and OpenAI API.

PythonLangChainFAISSBERT
Desktop and mobile layouts for the ProcureJs website.
2024

Designed and developed a responsive website for ProcureJs, focusing on user experience and functionality. Implemented SEO strategies that resulted in the website achieving #1 ranking in Google search, significantly enhancing online visibility and organic traffic.

SvelteTailwindJavaScript
Illustrative monitor frame with synthetic readings alongside the extracted values.
2023–2024

Developed an innovative workflow leveraging CCTV footage to extract vital information from patient monitors, optimizing nurse-to-patient ratio and cost efficiency with 90% accuracy of extracted data.

PythonYOLOOpenCV