I'm an Electronics & AI engineering student from Chalakudy, Kerala — obsessed with the point where firmware meets intelligence. From flashing microcontrollers at 3am to training models that run on 256KB of RAM, I build things that work in the real world.
My background spans mechanical fabrication (lathe, welding, CNC) and software systems (full-stack IoT, ML pipelines, WebSocket interfaces). I believe the best engineers are also craftspeople.
Currently pursuing B.Tech in CSE (AI) at Adi Shankara Institute of Engineering and Technology, while building projects that blur the line between hardware and software.
🔧 Hardware Hacker🤖 Edge AI Builder📡 IoT Architect🖨️ 3D Printing⚡ Firmware Dev🌐 Open Source
Digital Twin framework for precision farming. Builds a real-time edge-to-cloud IoT pipeline connecting ESP32 sensors to a Raspberry Pi hub and cloud analytics backend.
▸ESP32 → Raspberry Pi → Cloud data pipeline
▸Real-time MQTT telemetry with Flask admin & user dashboards
Smart Water Digital Twin system for ML-based pipe leak detection. Combines physical flow constraints with anomaly detection algorithms and live Streamlit dashboards.
4×4×4 LED cube driven by ESP8266 with a WebSocket-controlled web UI, real-time animation engine, and a Three.js 3D simulator for designing effects before flashing.
▸Real-time WebSocket animation control
▸Custom multiplexing driver in C++
▸20 animations, previewable in a Three.js browser simulator
Embedded arcade game running on ESP32-S3. A retro space shooter with procedural enemy waves, power-ups, and particle effects — built entirely from scratch on bare metal.
Modular robotics control framework built with PlatformIO and C++. Designed for scalable, reusable embedded robotics applications with clean hardware abstraction.
Decision-support digital twin for dam–river flood and hydropower optimisation on the Periyar basin, Kerala. Built for EVOKE 26 (Climate Resilience track): it replays the October 2021 Idukki flood and searches for the gate-release policy that should have been used.
▸Replays the Oct 2021 flood to within 0.30 m of the observed reservoir level
▸Optimiser finds ~3 m more flood cushion with slightly higher generation revenue
▸Holds up to 48 h ahead against a 30-member NOAA GEFS forecast ensemble
▸Advisory only — a named officer approves, the twin never operates a gate
Wearable closed-loop electrical muscle stimulation (EMS) physiotherapy platform for the hand. Electrodes on the back of the hand reach finger muscles that forearm-cuff devices miss; MIRA waits for the patient’s own EMG attempt, supplies only the missing effort, then fades as control returns.
▸Camera pose, EMG and IMU sensing; the AI plans movement but never sets a dose
The Uncooperative Waste Management System — a trash can that is technically intelligent and deliberately unhelpful. It opens for your rubbish, comments on your throwing and almost never obeys its industrial remote. A NORMAL MODE button works for about four seconds, then a mechanical finger switches it off.
▸ESP32-S3 controller with an ESP32-C3 ESP-NOW remote and a “mistranslation engine”
▸Web-remote build where the phone is the remote, speaker and eyes (ESP32-CAM)
▸Autonomous companion brain on Arduino UNO Q with a reflex-safety body controller
▸Spin-off: YOLOX-nano steers a bin under your throw, then slams the lid shut
Digital Operating Model for Residential Awareness — a Physical AI operating system for the home. A live digital twin that observes, plans with expected effects, acts, and verifies its own actions against independent physical evidence. Built for the Arduino Physical AI Challenge.
Can a microphone replace a rain gauge? Trains ML models to tell whether it is raining and how hard purely from the sound of rain, using a mechanical tipping-bucket gauge as ground truth.
▸Time-aligns raw .wav recordings with tipping-bucket readings
▸XGBoost, CNN, LSTM and Transformer models merged in a learned stacking ensemble
▸19 field recording campaigns (287.64 GB) published on Kaggle under CDOH Lab
▸Build guide and ARG Node v2 firmware for making your own
ESP8266 IR decoder with a live web dashboard. Point any remote at a TSOP1738 sensor and watch the protocol, hex code and bit count decode in your browser — no serial monitor needed.
▸Live decoding streamed over WebSockets using IRremoteESP8266
▸Name and save buttons to LittleFS, with a rolling signal log
▸Remote Learning Mode: photo a remote, map its buttons, capture each 3× verified, export CSV / JSON
▸Interactive browser simulator — try it with no hardware
ESP32 / ESP8266 LED chaser that hosts its own control interface. Pick from 53 patterns, design custom sequences with per-step timing, watch the browser mirror the strip, and download a ready-to-flash sketch for whatever you land on.
▸53 patterns, including every pattern of the original 13-LED sketch frame for frame
▸Sequence lab with per-step delays and JSON import / export
▸Simulator runs the same pattern engine as the firmware, plus a “Catch the Light” reflex game
▸Generates a ready-to-flash .ino for your configuration
A reaction game on a 24-LED WS2812B ring. A green cursor races round the circle — press the button exactly when it lands on the red target. The cursor speeds up as you score, all against a 60-second round clock.
▸One-button timing game with levels & a speed curve
▸Phone WiFi control panel — colours, effects, difficulty, live
▸6 LED effect styles, 11 buzzer sounds, 7 presets + save slots
▸SSD1306 OLED scoreboard with score, best, level & time bar
The classic memory game rebuilt for the ESP8266. Simon flashes a colour and plays a note — repeat the sequence, then one more step is added. Four glowing buttons, a 24-LED ring and an OLED scoreboard.
▸Sequences up to 50 steps, faster every level (600 → 180 ms)
▸4 buttons + 4 LEDs on one PCF8574 I²C expander — only 2 GPIOs
▸C · E · G · C note per colour on a passive buzzer
▸Faithful browser simulator of the real firmware logic
A physical reaction game: six sticks hang from six servo-driven hooks. After a countdown, the hooks let go one at a time in a random order at random intervals — catch them before they hit the floor.
▸6 SG90 servos driven by a PCA9685 16-channel PWM board
▸Fisher-Yates shuffle — every stick drops exactly once
▸Own WiFi AP with a mobile panel for timing & servo calibration
▸Non-blocking state machine keeps the web UI responsive
The original Cyclone mode: an arcade timing game on an ESP8266 with a 24-LED WS2812B ring. Tap the button when the running LED crosses the glowing target zone — six levels, each faster than the last.
▸6 progressive difficulty levels with a 6-LED gradient score strip (green to red)
▸WiFi AP web UI: RGB colour pickers, Easy-to-Insane presets and custom preset slots
▸Settings persist in EEPROM; rainbow attract mode while idle
▸Browser simulator with comet trail, sound, confetti on a win and screen-shake on a miss
Physical arcade games built on the ESP8266 and ESP32-S3 — LED rings, servos, buzzers, OLEDs and phone-friendly web panels. Each one is open source on GitHub with firmware and stand-alone hardware tests, and every one has a playable browser simulator.
A reaction game on a 24-LED WS2812B ring. A green cursor races round the circle — press the button exactly when it lands on the red target. The cursor speeds up as you score, all against a 60-second round clock.
▸One-button timing game with levels & a speed curve
▸Phone WiFi control panel — colours, effects, difficulty, live
▸6 LED effect styles, 11 buzzer sounds, 7 presets + save slots
▸SSD1306 OLED scoreboard with score, best, level & time bar
The classic memory game rebuilt for the ESP8266. Simon flashes a colour and plays a note — repeat the sequence, then one more step is added. Four glowing buttons, a 24-LED ring and an OLED scoreboard.
▸Sequences up to 50 steps, faster every level (600 → 180 ms)
▸4 buttons + 4 LEDs on one PCF8574 I²C expander — only 2 GPIOs
▸C · E · G · C note per colour on a passive buzzer
▸Faithful browser simulator of the real firmware logic
A physical reaction game: six sticks hang from six servo-driven hooks. After a countdown, the hooks let go one at a time in a random order at random intervals — catch them before they hit the floor.
▸6 SG90 servos driven by a PCA9685 16-channel PWM board
▸Fisher-Yates shuffle — every stick drops exactly once
▸Own WiFi AP with a mobile panel for timing & servo calibration
▸Non-blocking state machine keeps the web UI responsive
The original Cyclone mode: an arcade timing game on an ESP8266 with a 24-LED WS2812B ring. Tap the button when the running LED crosses the glowing target zone — six levels, each faster than the last.
▸6 progressive difficulty levels with a 6-LED gradient score strip (green to red)
▸WiFi AP web UI: RGB colour pickers, Easy-to-Insane presets and custom preset slots
▸Settings persist in EEPROM; rainbow attract mode while idle
▸Browser simulator with comet trail, sound, confetti on a win and screen-shake on a miss
Adi Shankara Institute of Engineering and Technology Focused on edge computing, embedded AI, and IoT systems design. Active in hardware labs, robotics club, and open-source firmware development.
2020 — 2023
Diploma in Mechanical Engineering
Thiagarajar Polytechnic College Hands-on training in lathe operation, welding, drilling, CNC machining, and mechanical fabrication. Built a strong foundation in physical manufacturing and precision tolerances.
// research interests
🧠Edge AI & TinyML
⚡Spiking Neural Networks
🔋Low-power IoT Systems
🌐Digital Twin Systems
📡LoRaWAN / Matter / Thread
📊Sensor Data Analytics
⏱️Real-time Embedded Systems
🤖Wireless IoT Protocols
06// passions
Things I Love Building
📡
Smart Connected Devices
Designing hardware that responds intelligently to its environment — from sensor fusion to cloud sync, making the physical world programmable.
🤖
AI on Microcontrollers
Embedding intelligence into constrained devices — TinyML, edge inference, and neural networks running on kilobytes of RAM with real-world latency.
🔧
DIY Electronics & Fabrication
Soldering iron in hand, schematic on screen — building from scratch, debugging signals, and the satisfaction of hardware that actually works.
🌍
Solving Real Problems
Technology that matters. Automation, monitoring, assistive devices — engineering with purpose, local context, and measurable impact.
