Projects I've built, prototyped, or actively explored. Each card names the actual problem area and links to a repository or live project where one is available.
STEVE is my ESP32-S3 robotics project. I've worked on microphone input, wake-word detection, a laptop-side speech pipeline, motor control, and a separate visual face server. The goal is to make the hardware, audio, and interface work together without adding unnecessary latency.
VIEW REPOSITORY ↗HARDWAREA Python/Tkinter desktop workspace combining AI chat, study features, a notes vault, MCQ generation, and image-generation integration. Firebase stores app data; I aim to keep the workflow useful without depending on a pile of paid APIs.
A field-ops app designed around the day-to-day needs of a pest-control team: attendance, technician questions, photo uploads, and digital staff IDs. The backend uses Supabase, with Google Drive for file storage.
VISIT LIVE SITE ↗FIELD OPSI publish experiments and project code on GitHub, use issues and version control to keep work organised, and try to leave repositories easier for another person to understand.
EXPLORE PROFILE ↗OPEN SOURCESmall machine-learning exercises focused on the full loop: preparing data, training a baseline, checking the result, and understanding where a model gets things wrong.
VIEW REPOSITORY ↗AI / MLA modular RAG project with document ingestion, OCR and audio transcription paths, FAISS retrieval, and optional web context. It is an experiment in connecting different input types to one retrieval workflow.
VIEW REPOSITORY ↗AI ENGINEERINGVantra is my lightweight Linux desktop exploration, based on Alpine Linux and XFCE. It reflects a recurring constraint in my projects: making tools usable on modest hardware instead of assuming a powerful machine.