Building AI systems for Vision, Language, and Trustworthy deployment.
I’m a Master’s student in Computer Science at the University of Massachusetts Amherst, with industry experience at Robert Bosch and research experience across Computer Vision, NLP and Agentic AI.
Interests: Machine Learning, Computer Vision, NLP, Generative AI, LLMs.
Education
Academics & coursework
University of Massachusetts Amherst
Relevant coursework: Advanced NLP; Robotics; Neural Networks: A Modern Introduction; Optimization in CS; Data Science Fundamentals; Research Methods for Empirical CS; Trustworthy & Responsible AI; Methods in Applied Statistics.
Indian Institute of Information Technology, Pune
Undergraduate foundation in Computer Science and Engineering.
Experience
Industry & research
Research Intern — Tech5 USA
- Built a facial occlusion detection and alignment framework for biometric identification using Vision Transformers and landmark-based models, achieving IoU 0.95, Dice/F1 0.97, Precision 0.97, Recall 0.97, and 97.9% pixel accuracy.
Data Scientist (Senior Engineer) — Bosch Global Software Technologies
- Contributed to three patents on Explainable AI for Bosch VivaRay : Anemia classification & hemoglobin prediction.
- Implemented end-to-end 3D object detection using LiDAR point clouds leveraging PointNet/PointNet++, VoxelNet, and PointPillars for autonomous driving scenarios.
- Built intelligent fault detection algorithms for predictive maintenance using time-series and electrical current data (anomaly detection + signal-based ML) to enable early failure prediction and reduced downtime.
- Developed an LLM-powered machine troubleshooting chat assistant using LangChain, enabling contextual reasoning, multi-step diagnostics, and retrieval-augmented responses.
- Developed Explainable AI algorithms (SHAP, LIME, Grad-CAM, etc.) for an in-house Responsible AI platform.
Intern — Bosch Global Software Technologies
- Built a framework for image and video dehazing using GANs and CNNs for diverse atmospheric conditions.
- Implemented real-time data augmentation to generate synthetic datasets with varying fog intensity levels.
Research Intern — University of Pavia, Italy
- Designed a multitemporal SAR feature selection strategy using seasonal and 24-day Sentinel-1 composites with spatial statistics to capture phenological and spatio-temporal patterns for vegetation classification.
- Implemented Random Forest land-cover mapping in Google Earth Engine.
MS Projects
Selected graduate work
Neural Networks — Video Crowd Density Estimation
- Adapted image-based crowd density estimation to video using diffusion-based denoising and event-driven optical-flow sampling for accurate real-time crowd monitoring.
Trustworthy & Responsible AI — Memory Poisoning Defenses (EHR Agents)
- Evaluated memory poisoning attacks in persistent-memory LLM agents for EHRs and introduced defenses, showing reduced attack effectiveness.
Advanced NLP — Medical VQA with KG + RAG
- Developed a medical VQA framework combining KG + RAG with multimodal retrieval and PrimeKG knowledge graphs.
- Evaluated on VLMs including LLaVA-RAD, CheXAgent, and Llama-3.2 Vision using ROUGE/BLEU and DeepEval.
Robotics — ROS Object-Tracking Robot (Triton)
- Deployed a ROS-based autonomous object-tracking robot using YOLOv3-tiny + SORT with color-histogram re-identification, achieving low-latency (~1.2s) real-time human tracking and autonomous following.
Publications
Links included
Human Face Generation from Textual Description via Style Mapping and Manipulation
Multimedia Tools and Applications (Springer), 2022
Mapping Globally Using Multitemporal Sentinel-1 SAR: A Semiautomatic Approach
IEEE InGARSS, 2021
Text-Guided Image Manipulation Using LiT and StyleGAN2
IEEE ICMI, 2024
XAI for Bispectrum-CNN Based Bearing Fault Detection
VETOMAC 2023; JVET (Springer)
XAI-Driven Spindle Health Monitoring for Precision CNC Machining
ICMER EC (Springer), 2024
Technical Skills
Tech stack & core skills