Faheem Thanveer

Architected an agentic PySpark ETL pipeline converting legacy C# scripts to PySpark with validation loops using LangGraph and Databricks. Leading a team for development and upkeep of the project.

Architected and deployed a full-stack video intelligence pipeline for ad-compliance auditing. Designed an end-to-end CCTV analytics system for warehouse audit and inventory estimation. Built and maintained an MCP-based enterprise agent, generating a new revenue stream. Led two teams of 20 engineers.

Studied complex pipelines using Nvidia Triton server for ingesting and processing Twitch or YouTube streams with detection and OCR. Curated new augmentation strategies improving accuracy from 88 to 93 for English events. Expanded to Spanish and Portuguese streamers. Trained a RegNet model for game identification.

Worked on a tool to process CAD drawings and find exact changes in revisions. Researched and deployed interactive segmentation models. Built a license plate detection and OCR model for CCTV footage.

Worked on 3D reconstruction and human pose estimation for animation. Developed a Python ML algorithm to ease animation studio workflow, reducing animation development time by 30 percent.

Initiated and assisted in capturing new metrics for evaluation. Built dashboards and SQL tables visualizing metrics for the CEO. Streamlined the SQL database and projected mutual fund data to predict investment performance.

Developed and managed an R-Shiny based sales dashboard for DXC Technologies. Designed ETL pipeline reducing load times by 70 percent. Developed a KYC automation algorithm using computer vision. Built a graph database and recommender system on Neo4j.

Masters in Computer Vision, Robotics and Machine Learning from the University of Surrey, United Kingdom.

Bachelor of Technology in Electrical and Electronics Engineering from the National Institute of Technology, Trichy, India.