Loading portfolio…
Data Scientist and AI/ML Engineer with experience building end-to-end ML pipelines, multimodal AI systems, ETL data infrastructure, and production dashboards across large-scale datasets. Proficient in Python, SQL, PyTorch, TensorFlow, and cloud platforms (AWS, GCP, Azure) for data engineering, model development, MLOps, LLM fine-tuning, computer vision, NLP, and BI reporting. Published researcher and cross-functional Agile collaborator. MS Data Science, GPA 3.7/4.0.
Xyton Semiconductors
Co-developed multimodal AI system (ViT, YOLO, ResNet + transformer LLMs) – fine-tuned foundation models using PyTorch and Roboflow, improving classification accuracy by 18%; built end-to-end netlist generation pipeline converting circuit diagram images to SPICE netlists, automating a previously manual EDA workflow.
Architected AWS ETL infrastructure (S3, Lambda, Glue) for LLM pre-training and fine-tuning across multimillion-record datasets – reduced processing time by 40%; implemented MLOps covering training, evaluation, deployment, and real-time inference monitoring; built data pipelines processing 500+ research documents via REST API.
BMW Manufacturing Co., LLC
Developed production routing dashboard using Oracle APEX, SQL, PL/SQL, and Oracle Database tracking 20+ critical KPIs for 300–500 vehicles/shift across Band F1B assembly line (H50 & H52) – enabling 10–30 workers and managers to identify correct test tracks and eliminate routing errors entirely.
Enhanced Manager’s Overview dashboard with dynamic color-coded KPI tracking and conditional formatting logic (under-target, at-target, over-processing) for road test and water test completion rates – eliminating 15–30 min/shift of manual reporting effort, enabling 3x faster bottleneck identification across 2 production facilities.
Indus University
Designed end-to-end ML pipeline on 100,000+ patient records (Python, scikit-learn, PyTorch, Keras) – feature engineering across 40+ clinical features; ensemble methods (Random Forest, XGBoost, SVM) achieving 92% accuracy across 5 model variants; reduced inference time by 45%.
Resolved 15+ failure modes across 3+ iteration cycles/month; curated 500+ training samples using Roboflow improving data quality by 30%; built Tableau dashboards for stakeholders; published in Web of Science (Feb 2024); mentored 3 junior team members.
Built multimodal AI pipeline integrating BLIP and Falcon LLM for multi-image understanding and narrative generation – applied prompt engineering, few-shot learning, and cross-modal embeddings; deployed as a full-stack production web application using Anvil with REST API endpoints and real-time inference.
Built end-to-end ML and ETL pipeline analyzing 50,000+ employee records – feature engineering across 30+ features, ensemble classification (Random Forest, XGBoost), and real-time Power BI dashboards; contributed to 5% attrition reduction and $200K+ annual savings.
University at Albany, State University of New York
Indus University, Ahmedabad