Skip to content

Autonomous Systems Developer

UditGarg — Autonomous Systems Developer building robotics, agentic workflows and self-hosted AI in New Delhi.

I build things that run on their own.

New Delhi·--:--:-- IST
Autonomous SystemsIndustrial RoboticsAgentic WorkflowsSelf-Hosted AIWorkflow AutomationEmbedded SystemsAutonomous SystemsIndustrial RoboticsAgentic WorkflowsSelf-Hosted AIWorkflow AutomationEmbedded Systems
(01)Statement

I build systems that keep working when nobody is watching: robots that take instructions in plain language, and AI that runs on hardware I own.

I would rather understand a whole stack than trust a black box, so almost everything here is self hosted: local models, my own vector store, my own queue. The interesting problems start once a thing has to keep running on its own.

A

Autonomous systems

Machines and pipelines that carry a task from start to finish without a person holding their hand at every step.

B

Agentic workflows

LLM driven orchestration with state validation, human approval where it matters, and a hard stop where it does not.

C

Self-hosted AI

Local models, custom quantisation and vector search running on hardware I control rather than someone else's API.

D

Industrial automation

Cobots, mobile robots and factory workflows connected to the software layer that actually decides what happens next.

(02)

Selected work

Also on GitHub
(03)

Night and altitude

A dark home office at night with three screens and magenta ambient light on the wall

Night

02:14 IST · the room

Snow covered Himalayan ridge above a pine forest

Altitude

2,050 M · the reset

The work happens in a dark room with one wall lit magenta. Everything that keeps it worth doing happens somewhere with no signal.

(04)

The desk

A dark home office at night with three screens, a backlit keyboard and magenta ambient light on the wall

The room

  • Comms rail

    Everything that interrupts, kept off the main screen.

  • The machine

    Editor, containers, and whatever model is running locally.

  • The reference

    Robot state on one side, running build on the other.

  • Where the hours go

    Most of this was built between midnight and four.

I got into this through hardware, then found out the interesting part was the software deciding what the hardware should do next. Now most of my time goes into the layer in between.

Day to day that is Python and ROS near the machines, Next.js where a human has to look at something, and a local model doing the reasoning in the middle. Almost none of it calls out to an API I do not own.

Studying at
Dr. Akhilesh Das Gupta Institute of Professional Studies
Based in
New Delhi, India

Tools in rotation

  • Python
  • TypeScript
  • FastAPI
  • Next.js
  • ROS
  • Ollama
  • LangGraph
  • CrewAI
  • Neo4j
  • Redis
  • Celery
  • Docker
  • n8n
  • Linux
  • AWS
  • ESP32
(05)

Trajectory

  1. 2023

    First commits

    Started building in public. Learned by taking systems apart rather than reading about them.

  2. 2025

    Hardware

    ElectroSentry: ESP32 corrosion monitoring with pH and TDS sensing, a dashboard and cathodic protection. Third at the college exhibition.

  3. 2025

    Models

    Grammar Score Engine. Speech recognition feeding supervised regression to grade spoken English without a marker.

  4. 2026

    Robotics internship

    IITD-AIA Foundation for Smart Manufacturing, through August 2026. Built natural language control for a Techman TM5M-700 cobot and an autonomous mobile robot.

  5. 2026

    Autonomy

    KhojAI. A self-hosted intelligence platform that reads fifty thousand documents a day so a person only has to read three hundred.