3 min read
Building PRISM in 24 hours at ASIET's AI Innovation Hackathon
I built PRISM, an evidence-first handwritten-paper assessment workspace, in 24 hours at ASIET and won second place, a cash prize, and an internship offer.
Midhun P M — BlogNotes from the workbench
What I'm building, what broke, and what I learned fixing it. No hot takes, no listicles.
4 min read
Taking Thursday to the OpenAI Codex Community Hackathon in Bengaluru
I took Thursday, my local-first Linux desktop assistant, to OpenAI's Bengaluru Codex Community Hackathon after selection from nearly 2,000 applications.
4 min read
Building an adaptive interview system at VicoDathon
I built Probe Interview for ABTalks' 48-hour VicoDathon sprint: a LangGraph system that turns learning history into grounded technical interview practice.
4 min read
How I built an AI agent for Kochi Metro in a 6-hour sprint
What it was like building MetroMind at OpenAI Codex Nightline — the world's first AI build sprint inside a moving metro. What we built, what broke, and what we learned.
4 min read
Running local LLMs on an Intel Arc GPU with Vulkan and LLaMA.cpp
Most LLM guides assume you have an NVIDIA GPU. I don't. Here's what actually works for running quantized 8B models on Intel Arc using Vulkan-accelerated LLaMA.cpp.