Debayan Sen

Sole engineer for the platform's AI services layer, designing and building the full pipeline end-to-end including data ingestion, automated database onboarding, process mining, AI reporting, and a tool-calling NL-to-SQL agent exposed over MCP. Achieved 83% text-to-SQL execution accuracy on an internal golden set, with LLM evaluation in DeepEval and observability in Langfuse.

Built and deployed a 3D point-cloud instance segmentation model (Mask3D, PyTorch) automating wall-insulation assessment, reaching 72.3% mAP @ IoU 0.5. Deployed a ROS-based real-time 3D perception system fusing LiDAR and camera streams for autonomous surveillance and robotics platforms.

Developed an end-to-end computer-vision pipeline for automated quality inspection, designing a weighted soft-voting ensemble of Mask R-CNN and U-Net segmentation models, reaching mean IoU 0.772 and Dice 81.9% on industrial cork-disk datasets.

Developed Qt/C++ industrial monitoring software for sensor-driven production environments and reduced image-transfer latency between embedded devices and industrial vision systems through optimised compression.

Developed and tested core LTE/5G network-simulation and mobility-management (MME) components in C/C++ and built backend and frontend functionality for telecom testing tools using C, SQLite, and Tcl/Tk.

AWS Certified Solutions Architect – Associate (SAA-C03) certification in progress.

Grade: 1.8

M.Sc. in Computer Science with final grade 1.8. Thesis on anomaly detection on cork disks, completed at Fraunhofer IPM with grade 1.3.

Grade: 8.43/10

B.Tech. in Information Technology with a grade of 8.43 out of 10.

Completed the Analyze Datasets and Train ML Models with AutoML certification.

Completed the Generative AI with Large Language Models certification.