How I fine-tuned Llama 3.2 3B to process transcripts locally, achieving better results than models 5-8x larger through specialized training.
Designing and building a distraction‑free Android app in Kotlin/Compose with low‑latency audio and a guided vocabulary mode.
Exploring a single-LoRA, tag-routed architecture that processes a Wikipedia article once and generates study guides and concept maps/timelines.
A plan to build a small local offline AI tutor for children in underserved communities.
Turning a 1B LLM into a deranged cosmic entity running on CPU only.