Tevan Bhatia

software engineer · new york city

I build backend systems and applied AI: LLM pipelines, agent orchestration, and the infrastructure that keeps them honest. The part I care about is the stretch where a model stops being a demo and starts being something a person can rely on.

tev@nyc: ~ — zshtype a command

tev@nyc:~$ help

available commands

whoami

now

skills

projects

background

fun

links

clear

↑ ↓ for history, tab to complete

Run a command

~/now

Founding engineer at Termblocks. I build agentic systems and LLM-driven document ingestion pipelines in Python. Most of my week goes to the unglamorous half of applied AI: making extraction reliable, making agent runs observable, and making the whole thing hold up in front of a customer.

~/skills

  • languagesPython · Go · TypeScript · SQL
  • aiLLM pipelines · agentic systems · PyTorch · RAG
  • backendFastAPI · Postgres · Prefect · Next.js
  • infraAWS · Terraform · Docker · Databricks

~/projects

  • ai integration firm

    An AI engineering practice built around LLM delivery, closing the gap between open model research and the systems that have to run it in production.

    Llama 2 · Next.js · Databricks · Terraform · AWS

  • e-commerce automation

    End-to-end automation of order placement, assignment and completion for a service business, with realtime updates and role-restricted views per user type.

    Laravel · Vue.js · MySQL · Node/Express · Pusher

  • parallel-harness-pets

    A creature for every git worktree, living in your coding agent's status line, so you can tell at a glance which parallel session has gone sideways.

    Go · Claude Code · tmux

~/background

Honours Bachelor of Science with High Distinction from the University of Toronto, specialising in computer science with a focus in artificial intelligence. I started in deep learning research, training CNNs for tissue classification, and moved steadily toward the systems side of the field.

~/off-the-clock

  • Bouldering, mostly the falling-off part.
  • Finding new restaurants, then over-recommending them.
  • Guitar — a Fender Telecaster and a Seagull S6.
  • Spotify playlists maintained with more rigour than some production systems.

~/elsewhere