Bridging the gap between intelligence and hardware — deploying smart models where resources are scarce.
I'm a robotics and electronics enthusiast passionate about making intelligent systems run on constrained hardware. My work sits at the intersection of embedded systems, computer vision, and edge AI — taking models that typically demand data centers and squeezing them onto microcontrollers and single-board computers.
class Sumit:
focus = ["Edge AI", "Robotics", "Embedded Systems", "IoT"]
hardware = ["Raspberry Pi", "ESP8266/ESP32", "Arduino", "NVIDIA Jetson"]
protocols = ["LoRa", "Zigbee", "UART", "MQTT", "I2C/SPI"]
currently = "Deploying intelligent models on low-performance boards 🔬"
ask_me = "Anything about robotics, edge ML, or embedded firmware"
| Project | Description | Tech |
|---|---|---|
| FedVibroSense_STM32 | A STM32F405RGT6 Based on-device federated learning node | C++ C STM32 Edge AI Fed_ML |
| SensiNerveX | On-device training of a feedforward neural network directly on the Seeed XIAO ESP32-S3 microcontroller | C C++ ESP32_S3 Edge AI Fed_ML |
| LoRa32T | An Encrypted LoRa telemetry over ESP32 With auto-adpative encryption mode | C++ ESP32 LoRa |
| LoRa8266T | LoRa telemetry over ESP8266 connected to MCUs via UART — long-range, low-power data link | C++ ESP8266 LoRa |
| smart_Air-pad | Turn any finger or pen into a virtual drawing tool — no touchscreen required | Python OpenCV Computer Vision |
| NanoSplat | A real-time pipeline that takes live monocular video as input and outputs a coloured, segmented 3D point cloud of a specific user-named object — on hardware that costs under $100 | Python OpenCV Computer Vision Jetson-Nano |
I maintain an ORCID researcher profile — my work explores practical applications of embedded intelligence in robotics and IoT systems.
- 🧩 Edge ML — compressing and deploying neural networks on microcontrollers and SBCs
- 📡 LoRa & Zigbee mesh networks for sensor telemetry
- 🦾 ROS-based robotic systems with real-time vision pipelines
- 🏠 Home automation systems with local-first AI inference
"The best robot is one that thinks for itself — even on a $5 board."
⚡ Let's build something intelligent together.