About Me
Senior IoT & Edge AI Engineer
Senior IoT and edge AI engineer with 7+ years of experience shipping products in healthcare, wearables, and fitness tech. I work on on-device ML (EfficientNet-B0, TFLite, ONNX Runtime, INT8 quantization) and embedded firmware (nRF52, ESP32, FreeRTOS, BLE/GATT). I led an 8-engineer team building a medical device platform that passed regulatory review on its first submission. My work spans the US, UK, Belgium, and India.
- EfficientNet-B0 to ONNX INT8: 94.7% accuracy, <500ms on Pi Zero 2W
- TFLite on nRF52: 92% accuracy, <50ms inference on Cortex-M
- Dual-hardware IoT: Pi 5 and Pi Zero 2W with AES-256-GCM BLE and AWS IoT
- 40% faster releases through CI/CD, with 500+ active patient sessions
Experience
Senior Engineer & Technical Lead
OraLens Healthcare · Bengaluru, India
- Led an 8-engineer team building a cloud-connected medical device platform. The architecture passed regulatory review on its first submission.
- Designed a dual-hardware IoT system (ESP32 plus Raspberry Pi 5 and Pi Zero 2W) with AES-256-GCM encrypted BLE, MQTT sync through AWS IoT Core, and a 72-hour SQLCipher offline buffer. The system supports 500+ active patient sessions in production.
- Trained and deployed EfficientNet-B0 with ONNX Runtime and DirectML. It reached 94.7% accuracy across 11 oral disease categories on 78,000 images, reducing clinician review time by 30%.
- Applied INT8 ONNX quantization for a 2 to 3x inference speedup with less than 2% accuracy loss. The Pi Zero 2W deployment uses 1.5 to 2W, 10 to 20% CPU, and under 280MB RAM.
- Built CI/CD pipelines and automated test frameworks, making releases 40% faster and cutting post-release defects by 25%.
IoTHardwareML/AIAWSCI/CD
Jan 2022 to Jan 2025
- Designed wearable firmware from scratch on nRF52, covering the BLE/GATT stack, sensor integration, power management, and on-device DSP. It shipped to the EU market and was used in 30 to 50 care homes monitoring about 3,000 elderly residents.
- Deployed a TFLite activity-recognition model through Edge Impulse. It reached 92% accuracy with under 50ms inference on a Cortex-M MCU.
- Built adaptive sensor duty cycling and dynamic sampling that extended battery life by 40%, or about two days between charges.
- Built a sensor-data pipeline for collection, labelling, and preprocessing, cutting ML model iteration time from weeks to days.
- Set up HIL, unit, and integration testing. Release quality improved by 35%, and QA cycles fell from two weeks to three days.
BLEFreeRTOSEdge MLFirmwareEU Market
IoT Systems Engineer, Fitness Technology
Renovatio Systems Ltd· United Kingdom
- Built ESP32 firmware for smart gym equipment with real-time strain-gauge load measurement, accurate to ±0.5 kg across 0 to 200 kg.
- Added MQTT cloud sync for live workout tracking and real-time data delivery to 1,000+ users.
- Tested four sensor technologies (strain gauge, load cell, piezo, and capacitive) and recommended the best balance of accuracy, cost, and power.
- Produced manufacturing-ready deliverables including circuit schematics, system architecture docs, and calibration procedures.
ESP32MQTTFirmwareSensor Integration
Project Engineer
Ralph L Wadsworth Construction· Draper, Utah, USA
- Led engineering work on the Denver Airport expansion and delivered two months early through careful sequencing and quality control.
- Set up QA inspection protocols and design-change workflows. The project had zero safety non-conformances.
Project ManagementQAConstruction Engineering
Technical Skills
Embedded & Firmware
ESP32STM32Raspberry Pi 5/Zero 2WnRF52FreeRTOSPlatformIO
IoT & Protocols
BLE / GATTMQTTI2C / SPI / UARTAWS IoT CoreAES-256-GCMSQLCipher
ML & Edge AI
ONNX RuntimeDirectMLEfficientNet-B0INT8 QuantizationTensorFlow LiteEdge Impulse
Robotics & Control
ROSSLAMPID ControlPath PlanningMATLAB/SimulinkImpedance Control
Sensor Fusion & DSP
IMUStrain GaugeLoad CellKalman FilterReal-Time DSPSignal Processing
Full-Stack & Cloud
Go / GinReact / Next.jsFlutterPostgreSQLDockerAWS (EC2, S3, Lambda)
Education
M.S. Mechanical Engineering, Robotics & Control Systems
University of Utah
Salt Lake City, UT, USA · 2019
GPA 3.6 / 4.0B.Tech Mechatronics Engineering
S.R.M University
Chennai, India · 2012
GPA 8.8 / 10Certifications
Engineering and product
Build across the stack.
Good products benefit from people who can move between firmware and the feature spec. BLE stack latency affects the user experience, and a clinical workflow constraint can change the data model.