Case Studies
13+ years of embedded engineering — from hospital AI robots to industrial weighbridge automation. Every project below is production work delivered for a real client.
Hospital Edge AI Patient Monitoring & Autonomous Robot
Fall detected and nurse alerted in under 3 seconds — fully offline
Built a privacy-first hospital monitoring system that detects patient falls using YOLOv8, retrieves medical context from a local ChromaDB vector database, reasons with a custom 135 MB edge-fine-tuned LLM running via Ollama, and dispatches an autonomous ROS2/LiDAR robot — all on Raspberry Pi 4 with zero cloud dependency. Federated Learning across edge nodes improves the model without ever sharing raw patient data.
Tech Stack
EdgeScribe — AI Conversation Analysis & Coaching Platform
Live product: auto-transcribe any professional conversation + AI quality scoring + Google Sheets report
Built a full-stack AI platform that records doctor-patient (or any professional) conversations, automatically transcribes them using speech-to-text, scores communication quality using AI, and delivers instant reports to supervisors via email and Google Sheets. Designed as a multi-industry product — the same platform works for healthcare, sales coaching, education, and HR interview feedback with zero reconfiguration.
Stack
Truck Weighbridge Alignment & RFID Identification System
Zero incorrect weight readings via 6-point alignment + fail-safe logic
Automated a truck weighbridge using UHF RFID vehicle detection, 6-position sensor alignment verification, and dual RS485 communication with PC weighing software. Fail-safe logic blocks weight measurement until the truck is correctly positioned, then triggers audio guidance, takes the reading, and releases the boom barrier automatically.
Stack
Industrial Water Level Monitoring via GSM & MQTT
700+ lines of legacy firmware rescued — deployed in 3 months
Took over a failing client project for a Bhopal startup. Debugged and restructured 700+ lines of unclear ATmega323 legacy code, rebuilt bidirectional MQTT communication over SIM A7670C 4G GSM, calibrated a 26 GHz radar level sensor using a 16-bit ADC, and delivered a production IoT system with real-time cloud dashboard, remote command handling, and EEPROM-backed local storage.
Stack
Multi-Network IoT Device with Automatic Communication Failover
Near-zero downtime — auto-failover across WiFi, Ethernet, and 4G
Built an industrial IoT device that seamlessly switches between Ethernet, WiFi, and 4G GSM based on real-time network availability. The device reads sensor data, creates structured payloads, and transmits continuously using whichever network is live. Includes remote relay control for equipment management — designed for industrial sites where any single network can go down.
Stack
Have a similar project in mind?
From hospital AI to industrial automation — if it runs on hardware and needs intelligence, we build it.