Lead Data Scientist - APAC

Qorvo

Staff Data Scientist

Infineon Technologies

• Developed & deployed scalable AI/ML models for Manufacturing/Engineering, including Wafermap Image Classification (ResNet50), Rule Optimization (xgBoost), and Failure Prediction (RUSBoost). These models led to substantial harvestings of ~420K€/year
• Deployment was done using FastAPI/Streamlit/SeldonCore on Openshift PaaS in K8S PODs. Utilized GitHub Actions & webhooks for CI/CD. The performance of the model monitored using Tableau Dashboards

• Developed Data extraction & Analysis scripts for Automated Lot-on-hold Analysis (100+ lots/hour over 6 sites across the globe) using MS-SQL,R & UiPath RPA

• Developed datasheet Q&A Chatbot using Llama2-7B-GGMU. Utilized LangChain to vectorize docs & store to FAISS VectorDB

• Experience in using Elastic Search, Filebeat, & Logstash to process large amounts of log data and generate insights on Kibana dashboard, resulting in improved debugging performance. Also store the processed data using Datalake API

• Developed a diverse set of advanced models using TensorFlow, Scikit-learn, pandas, pySpark, and Keras libraries. Models included CNN, SVM, Clustering, Classification/Regression using DecisionTrees/RandomForest /xgBoost. This comprehensive skill set empowered accurate data-driven insights

Senior Data Scientist

Qualcomm

• Developed an image classification model using YoLov3 for Failure Analysis image segmentation & fail-chip classification. Achieved a segmentation accuracy of 99% & a fail-chip classification accuracy of 93%
• Utilized GAN to improve the resolution of low-quality images from an old camera, resulting in high-res images comparable to those from a hi-res camera
• Expert in Statistical Data analysis & visualization of production test data using Python, NI O+, Spotfire(Exensio) & Tableau
• Experienced in applying Machine Learning models like Regression, Clustering, xGBoost & random forest for data interpretation using Keras
• Managed a team of 5 engineers handling Characterizing for multiple Mixed-Signal IPs in the SoC
• Development of end to end test automation framework for PVT char using LabVIEW & Python Framework
• Post-Silicon Validation of Mixed-Signal IPs on State of the art Mobile, Modem (5FF) & RF(14nm) SoCs
• Design & Development of High-Speed (>15Gbps) System Level Test Platform for Validation of 5G (Sub6/mm Wave) RF transceivers

Research Student (AI/Machine Learning)

ST Engineering & NTU Corps Lab

• Developed python UI for annotating features on airplane door images (viz door, door handle, window etc) , facilitating precise labelling for subsequent analysis
• Employed deep learning techniques (CNN using tensorflow1.0) for object detection, utilizing annotated data to identify and locate features on airplane doors with 95% accuracy
• Using the detected features Validating & the laser distance sensor, developed a complex scoring algorithm to confirm if the robot base is facing the airplane door
• Control the robotic base carrying the camera, laser sensor and aero-bridge depending on the control signals obtained, thus achieving a perfect docking (using Robotic Operating System)

Engineering Intern

Qualcomm

Validation Engineer

Texas Instruments