AI · Embedded Systems · Machine Learning
Building meaningful AI and software projects, learning continuously, and documenting the journey one step at a time.
01 — About
I'm a recent 12th-grade student from Indore, starting my B.Tech in Computer Science Engineering in August 2026.
My college hasn't started yet, and I'm already knee-deep in Edge AI and embedded systems — flashing boards, breaking builds, and figuring out how to fix them again. My main hardware is an NVIDIA Jetson Orin Nano 8GB, given to me by my father, which has become the base for everything I build.
I document the real process — including the errors most tutorials leave out — on YouTube and LinkedIn, while I keep learning Python, Linux, and how LLMs actually work under the hood.
2026B.Tech CSE begins
8GBJetson Orin Nano
100%self-taught
02 — Skills
03 — Projects
A locally-hosted chat application running Llama 3.2 3B on-device — llama.cpp serving the model, FastAPI backend, MongoDB Atlas for chat history, React frontend, exposed via Cloudflare Tunnel.
llama.cpp · FastAPI · MongoDB · React · CUDA
Full CLI flashing workflow for the Jetson Orin Nano, including a board-specific external device fix, plus recovery from a broken GPU driver state after a bad system upgrade.
L4T · JetPack · Bash · Linux
Ran Stable Diffusion Turbo directly on the Jetson, generating images in roughly 3.7 seconds after resolving a GPU driver bug that had broken the board.
Stable Diffusion · CUDA · Python
Set up and tested YOLOv8 for real-time object detection on Jetson hardware as part of exploring on-device computer vision.
YOLOv8 · Python · CUDA
A fully offline voice assistant running speech-to-text and an LLM locally on the Jetson — no cloud calls, everything on the edge.
Whisper · llama.cpp · Python
A written companion to my videos — deep dives into the errors, fixes, and edge-AI experiments that don't fit in a short video.
Writing · Edge AI · Linux
04 — Journey
Started from zero on Linux — file systems, the terminal, and how a real development environment works.
Documented the full CLI flashing workflow, including a board-specific device fix that most tutorials skip.
Fixed a GPU driver bug after a bad system upgrade — then ran Stable Diffusion Turbo generating images in ~3.7s.
Currently running Llama 3.2 3B on-device with a FastAPI backend, MongoDB, and a React frontend.
05 — Connect