Veerendra R. Patil
Bengaluru, India — open to relocation
AI× Robotics× Computer Vision× Embedded Systems

Veerendra R. Patil

I started building robots in 7th grade and never stopped — the machines just got harder. Today that means drones, competition robots, and neural networks compressed small enough to run on a microcontroller.

From a PyTorch model to real-time inference on an ESP32-S3

  • EducationB.Tech CSE, NIT Durgapur · CGPA 9.17
  • CurrentlyAI Solution Architect, Nautomation Labs
  • LeadingCCA RoboCell Head
Explore work↓ GitHub↗ LinkedIn↗ Résumé↓
8 years
of building
machines

It started in 7th grade,
with a board full of LEDs.

Sparky came out of that first stretch of tinkering — a board that did something on its own. Everything since has been the same instinct with harder problems attached: competition robots, drones, and models small enough to run on a microcontroller.

Sparky — an early buildWhere it started

What I build

Four domains — one problem

01

Robotics

Robot software and embedded systems — firmware, control loops and the plumbing that keeps a machine responsive.

02

Edge AI

Taking trained models through export, quantisation and deployment until they run inside a few hundred kilobytes.

03

Computer Vision

Real-time detection on live video, and turning pixels into calibrated real-world measurements.

04

Autonomous Systems

Drones and ground robots — telemetry links, sensing, flight control and the software you steer them with.

Selected work

Six projects — ordered by depth, not date

Training is the easy half

Flagship — deployment path

A model that only exists in a notebook has not been deployed. The Edge AI activity-recognition project takes six trained architectures the whole distance — through export, graph rewriting, INT8 quantisation and firmware — until they run on a microcontroller against live sensor data.

Deployment path: PyTorch to ONNX to TensorFlow to TFLite Micro to ESP32-S3 PyTorch CNN · CNN-LSTM · CNN-GRU Optuna search · CUDA/AMP ONNX Graph export Unrolled-RNN wrappers TensorFlow Conversion target WHILE ops eliminated TFLite Micro INT8 quantised <1e-4 output deviation ESP32-S3 C++ firmware Live IMU · on-device benchmark
16human activities classified
85.6%best benchmarked accuracy
6quantised models on-device
5×fewer parameters in the deployed model
16 MBcustom dual-slot OTA partition

Figures as reported in the project résumé — see the case study for method

Read the case study↗

Built, flown,
and benchmarked

Hardware — from the workshop

The Robocon competition robot on an outdoor basketball court with a ball loaded into its launcher
Robocon 2025 robot — on the court01
An ESP32-S3 watch running a quantised CNN on live IMU data, showing the prediction, confidence and inference time
Edge AI — live on-device inference02
A custom ESP32-S3 quadcopter flight-controller board held in one hand
ESP32-S3 flight controller03
Holonomic drive test04
Night flight test05
Legged walker — gait test06
Front view of the Robocon robot showing the launcher wheels and drive electronics
Launcher stage and drive electronics07
Aeris GCS — live telemetry08
Top-down view of a quadcopter airframe with motors, ESCs and wiring
Airframe09
The watch showing an on-device model picker with six tiles: left and right CNN, LSTM and GRU
Six quantised models, switchable on-device10
Rim detection + calibrated distance11

Experience

Shipping to production

May 2026 — Present Bengaluru, India

AI Solution Architect & Technology Consultant

Nautomation Labs

  • Architected and built a full-scale ERP system for industrial clients — a lean, cost-effective alternative to legacy platforms like SAP.
  • Designed and built decibyl.ai, the company's flagship GenAI voice-agent platform, orchestrating voice campaigns, lead qualification and follow-ups end-to-end through n8n workflow automation.
  • Single-handedly owned decibyl.ai's path to production: containerised it with Docker and shipped it live on AWS through a CI/CD release pipeline — GitHub Actions → ECR → EC2.

Contributor to
technical lead

RoboCell — CCA, NIT Durgapur

2024
Junior Member Embedded systems and robotics build work.
2025
Senior Member Software and embedded development on competition robots.
2026
CCA RoboCell Head Leads software and embedded systems development for the club.
Tooling
Robozido

Leads software and embedded development for the club, shipping in-house tools — including a Flutter + ESP8266 robot control app used across builds.

Funding & outreach
INR 8L+

Raised through the Robocon 2024 sponsorship campaign and multi-college skill-development workshops — funding and revenue brought into the club.

Competition
Robocon 2025
Finalist

Reached the final at IIT Delhi, leading embedded and software development and building the rim-detection vision model for the competition robot.

In-house tooling

RoboCell runs on software the club writes itself. Robozido is the control app the team reaches for — dual joysticks, servo and arm control, talking to an ESP8266 over Wi-Fi — so a new build can be driven the day it moves.

The Robozido Flutter app showing a dual-joystick manual control screen for a robotic arm
Robozido — manual controlFlutter + ESP8266
A row of identical student-built line-following robots lined up on a stage at a RoboCell workshop
Workshop build — one robot per participant01
In-house control app, driving a build02
Workshop components laid out and counted: jumper wires, motors, sensors, glue guns and chargers
Kitted and counted, per participant03
The RoboCell lab with members working at benches
RoboCell lab04

Achievements

Selected

2025 — IIT Delhi
Robocon
Finalist

Led embedded and software development; built the rim-detection vision model for the competition robot.

2023 — JEE Mains
99.23%ile

All-India engineering entrance examination.

2023 — KCET
Rank 66

Karnataka Common Entrance Test.

2023 — 2027
CGPA 9.17

B.Tech Computer Science & Engineering, NIT Durgapur.

Toolkit

What I actually reach for

AI & Machine Learning

  • PyTorch
  • TensorFlow / Keras
  • Scikit-learn
  • YOLOv5
  • OpenCV
  • CUDA
  • Optuna
  • ONNX
  • TFLite Micro

Robotics & Embedded

  • C++
  • C
  • ESP32 / ESP32-S3
  • ESP8266
  • Arduino
  • PlatformIO
  • IMU processing
  • PID control
  • CRTP
  • OTA updates
  • WebSockets

Backend & Cloud

  • FastAPI
  • REST APIs
  • Docker
  • AWS EC2 / ECR / S3
  • MLflow
  • CI/CD
  • GitHub Actions
  • MongoDB
  • MySQL

Software

  • Python
  • PySide6
  • Flutter / Dart
  • JavaScript
  • SQL
  • Git
  • Linux

Résumé

Two versions — same work, different emphasis

One page each. The robotics version leads with embedded systems and autonomous hardware; the AI/ML version leads with the model and infrastructure work. Education, experience and achievements are identical in both.

Download — Robotics↓ Download — AI / ML↓

Education

National Institute of Technology Durgapur

B.Tech, Computer Science & Engineering · CGPA 9.17 / 10 · Aug 2023 — May 2027 (expected)

Selected learning — Deep Learning & NLP · Computer Vision with OpenCV · Raspberry Pi & Arduino · C++ Data Structures

Veerendra R. Patil crouched beside the Robocon competition robot in the RoboCell lab

Veerendra R. Patil

Eight years of building machines that have to work in a room, not in a notebook.

Computer science at NIT Durgapur, embedded and software lead at RoboCell, and an AI solution architect shipping production systems at Nautomation Labs. The through-line is the same everywhere: get the model out of the notebook and onto something that moves.

Let's build
intelligent machines.

Eight years of building machines

I'm interested in robotics, computer vision, embedded AI, autonomous systems and the kind of engineering problems that only show up once software meets hardware. Based in Bengaluru, open to relocating.

Elsewhere

GitHub — veeru413↗ LinkedIn↗ Résumé — Robotics↓ Résumé — AI / ML↓
Veerendra R. Patil — © 2026 AI × Robotics × Embedded Systems