A machine learning and embedded robotics engineer who enjoys building real world AI systems that stay stable after deployment. My work sits at the intersection of perception, edge reliability, and production pipelines. I take ownership of calibration, evidence workflows, and system health in long running deployments, not just model accuracy.
At ESSI Integrated Technologies (New Delhi), I built a production grade radar + camera speed enforcement system using TI mmWave radar, delivering multi object tracking (up to 20 targets) with ~50 ms decision latency in roadside trials. I also developed a Flask + JavaScript dashboard for monitoring, analytics, and violation review, and hardened the pipeline with parsing, point cloud processing, Doppler analysis, and multi firmware handling across multiple hardware variants. To keep it field ready, I added watchdogs and automated recovery, achieving ~95% uptime with ~2 minute average recovery, plus radar camera calibration and PTZ auto slew that reached ~90% target lock success under 2 seconds.
Alongside this, I worked on high throughput ANPR pipelines supporting ~1600 cameras statewide, using Kafka for transport and Redis for state/caching to deliver end to end alert generation under 10 seconds. I also helped design resilient microservices with automatic restarts and RTSP reconnection cycles, and contributed to a scalable multi camera analytics architecture using FastAPI + Docker with modular services designed for independent scaling.
My toolkit includes Python, C/C++, Embedded C, backend frameworks like FastAPI and Flask, and infra like Kafka, Redis, Celery, PostgreSQL, Docker, Git. On the ML side, I use PyTorch, TensorFlow, Scikit Learn, OpenCV, NumPy, Pandas, and I also explore agentic AI workflows with Hugging Face, LangChain, and RAG.
I’m currently pursuing a BTech in Robotics and Automation at Symbiosis Institute of Technology, Pune, and I keep building robotics projects like ROS based object detection on Turtlebot 4, an ESP32 hydroponics monitoring system, and a bio inspired quadruped.
A crab-inspired quadruped robot using 18 MG996R servos, Raspberry Pi, ESP32, dual robotic arms, LIDAR, IMU, and a Pi Camera.
Features:
- Obstacle avoidance and autonomous navigation using ROS2
- Object detection and classification using computer vision
- Gesture and speech control interface
- Real-time camera feed and custom-built feedback UI
Designed a hydroponic automation system with pH, DHT11, LDR, and soil moisture sensors integrated via ESP32, Firebase, and ThingSpeak.
- Built an LSTM model to predict irrigation needs from time-series data
- Achieved high prediction accuracy and optimized water usage
An AI-powered system that recognizes 20+ exercises, counts reps, and provides real-time feedback.
- Utilizes OpenPose and pose estimation for accurate classification
- Live video feed with computer vision-based feedback system
