Niels Erik Nandal
Engineer building AI products, systems, and readable software.
I work across Python, TypeScript, and ML-heavy product surfaces. I care about deterministic backends, sharp product loops, and code a maintainer can understand in one pass.
I am currently open to full-time engineering roles.
featured projects
Selected work that best reflects how I think, what I ship, and the level of technical depth I want to be judged on.
niels-gptEnd-to-end LLM pipeline from tokenizer to chat interface, laptop-first
NexusReading-first knowledge management with semantic search and persistent anchoring
Factory SimulatorMulti-agent LLM orchestration with deterministic factory scheduling simulation
Suno Session Lab: turning clip generation into a navigable spaceTurns music generation into a navigable space, layering LLM labeling over CLAP/embedding-driven clustering and centroid-guided regeneration
writing
Technical notes and build writeups. Short, direct, and focused on the work.
- Zero to Hero - One
A plain-English walkthrough of embeddings, hidden layers, logits, loss, and backprop.
- Zero to Hero - Two
Why bad initialization breaks learning, and how Kaiming init and batch normalization keep the signal alive.
- My First Project
What building my first real full-stack app taught me about decomposition, realtime state, and avoiding style-first work.