bc

ML/AI Engineer

Bhuvan
Chennoju

Good at making machines learn. Better at making them not fail. Writing about ML, agents, and building things that actually work in production.

now

Day job: fraud ML, LLM agents, and payment optimization at T-Mobile. Nights and weekends: writing what I learn, experimenting with Claude Code, and reading everything I can about how models actually run at inference time.

writing
all →
GenLMcompleted

Genomic language model pretrained from scratch on 13 Drosophila species genomes — 12-layer dilated CNN with a custom character-level tokenizer, trained to predict variant effects across the genome.

Local RAGcompleted

Barebones local retrieval-augmented generation pipeline — PDFs in, answers out, with no API calls leaving the machine. Uses Ollama, ChromaDB, and LangChain.

ChottaLLMcompleted

Small-scale LLM built from scratch in PyTorch — character-level tokenizer, transformer architecture, trained on custom text corpora.

GPT-style autoregressive language model implemented in PyTorch — full training loop, attention, positional encoding, and character-level generation.