Quality > Quantity: Building DarkForest
How I designed a custom 128K Byte-BPE tokenizer for DarkForest, and the data curation strategy that turned a noisy corpus into something a model can learn from.
#llm#machine-learning#tokenization#bpe
Abhishek Gahlot
Senior engineer writing about infrastructure, distributed systems, and AI research. Currently training LLMs from scratch and building things that scale.
Writing
07 postsTraining a 7B LLM From Scratch
No fine-tuning. No LoRA. No base model. Training a 7 billion parameter language model from random weights, on rented H100s. General-purpose first, then specialized for code through SFT and RL on real agent trajectories.
#llm#machine-learning#training#from-scratch
Multi-Agent Orchestration for AI Chat Systems
How to build a chat system where multiple specialized AI agents handle different types of requests instead of one agent trying to do everything.
#ai#agents#architecture#typescript
Streaming Long-Running AI Operations with SSE
When your AI operation takes 30+ seconds, you cant just show a spinner. Here is how to use Server-Sent Events to stream progress in real time.
#ai#streaming#sse#architecture
Building Native Agent Team Mode for Codex
How I added Claude Code-style team coordination to Codex — one lead agent spawning and coordinating multiple teammates in parallel using a shared filesystem task board.
#ai#agents#codex#open-source
Building a Chat UI Component Library with React
How I built a reusable chat component library with React 19 and Vite that works across multiple projects.
#react#ui#vite#typescript
Hello World
First post on abhishek.it.
#meta