Loading...
AI / ML Embedded Software Engineer
Building and deploying intelligent systems onto real hardware — from edge AI and drone swarms to NLP and large language models.
AI / ML Embedded Software Engineer
Little Place Labs
Engineering systems that deploy intelligence onto real hardware.
Heterogeneous two-drone swarm (Scout quadcopter + Mule coaxial octocopter) for autonomous search and rescue. Jetson companion computers communicate with Pixhawk flight controllers via MAVLink/UART for real-time human detection, GPS geotagging, and autonomous payload delivery with dual verification protocol.
Impact
Enables fully autonomous search and rescue with dual verification — GPS coordinates plus secondary visual confirmation — eliminating false drops.
Results
Real-time human detection at 30 FPS with custom GNC stack and resilient 4G/WiFi telemetry bridge for BVLOS operation.
A locally-deployed, fine-tuned language model that learns and mirrors an individual's communication patterns, knowledge base, and decision-making style. Built for privacy-preserving personal AI that runs entirely on consumer hardware.
Impact
Trained on 3.4M tokens (1.9GB) of personal technical and reasoning data — achieves coherent persona mirroring.
Results
Fine-tuned for 137+ hours with loss 0.35, merged and quantized to GGUF for offline CPU deployment.
AI-powered drone platform for infrastructure inspection. Deploys optimized computer vision models on Jetson edge devices for real-time crack detection and solar panel assessment with TensorRT acceleration.
Impact
Enables on-device defect detection at the edge for critical infrastructure monitoring without cloud dependency.
Results
Achieves 99.3% detection accuracy with 250ms median latency using TensorRT and INT8 quantization.
Peer-reviewed research in edge-AI and autonomous systems.
Kumar Satyam, Adrish Haldar, Sanket Mondal, Sudip Santra, Chancharika Das, Jagriti Budhiya, Suranjit Ghosh, Soumik Dey, Sayan Pal, Abhro Mukherjee, Mourani Sinha · IEEE (Under NMICPS) · 2025
Proposes a heterogeneous drone swarm architecture combining agile Scout and heavy-lift Mule platforms with decentralized edge AI on NVIDIA Jetson Orin Nano Super for real-time human detection, autonomous decision-making, and closed-loop control. Introduces a dual verification protocol and secure beyond-line-of-sight communication framework using 4G-LTE and Cloudflare Zero Trust Tunnels.
Recognition from research institutions, hackathons, and engineering competitions.
Granted Design Patent
AI-Based Modular Crack Detection Drone System (10 Nov 2025)
IIT Kharagpur Incubation Pitch
Selected to pitch at IIT Kharagpur Incubation Centre
Technothon IoT Winner
First place in IoT-focused national hackathon
AI Unleashed Runner-Up
Second place in AI/ML competition
SIH 2024 Institutional Finalist
Smart India Hackathon institutional finalist
LPL Engineering Hackathon Winner
3rd Prize in internal engineering hackathon at Little Place Labs
B.Tech – Computer Science (AI/ML)
Computer Science with specialization in AI and Machine Learning
Techno India University, Kolkata · 2025
CGPA: 8.0 / 10
Engineering AI systems that run in the real world. Open to research, collaboration, and production deployments.
Repositories updated in the last 7 days
Repo for connecting and running inference on esp32-s3 cam board
"Am I audible?" Solved. A Linux CLI tool to seamlessly record, mix, and transcribe system and mic audio with zero dropouts.
A cli based tool that lets you share your computer's screen and one external monitor via wifi, using local host.
Local agentic coding platform, runs models via ollama.
Recent GitHub activity