Sriharsha Mopidevi

University of Pennsylvania, AI-4-AI Lab · Philadelphia, PA

• Designed and built MedVidDeID, a modular medical audio-video de-identification pipeline using Python, OpenCV, pose estimation, and ASR models. Achieved 92% de-identification accuracy and reduced processing time by 63% over manual methods.
• Developed and validated the pipeline through multiple peer-reviewed publications and conference presentations.
• Built and optimized Databricks workflows to process and query 10TB+ of multimodal data in real time, enabling concurrent experiments across multiple labs.
• Developing a multimodal FAIR data repository and dashboard using Azure Data Lake, Next.js, Django, and MariaDB.
• Developed HPC scripts using SLURM and LSF, reducing experiment runtime by 20%.
• Enhanced REDCap functionality with AWS and Amazon Mechanical Turk, enabling scalable video data labeling with 350+ valid responses.
• Led training sessions for junior developers and researchers on OpenCV, Docker, HPC workflows, and Databricks.
• Collaborated across labs to support NIH grant proposals, securing funding for AI-driven bioinformatics projects.
• Co-authored 6 peer-reviewed papers including publications in JBI, JAMIA, AAAI, and PSB, covering video de-identification, multimodal datasets, VQA benchmarks, and speaker role identification.
• Built containerized ML inference pipelines with Docker for reproducible deployment of computer vision and NLP models across HPC and cloud environments.
• Designed and implemented REST APIs with Django and FastAPI to serve multimodal clinical data to frontend dashboards and external research tools.
• Developed automated data quality checks and validation pipelines for clinical video and audio datasets, ensuring compliance with IRB and HIPAA standards.

GSPAN · New Delhi, India

• Developed a User Management System PWA using React.js, Python, and Django, increasing organizational productivity by 25%.
• Built an interactive web app visualizing Indian election and COVID-19 data using React.js, Deckgl, Mapbox, and React-vis, boosting website traffic by 5x.
• Migrated from monolithic architecture to microservices with Node.js and Express.js, reducing development cycles by 50%.
• Implemented CI/CD pipelines using Jenkins and GitLab, reducing deployment errors by 70%.
• Built a serverless architecture with Netlify and AWS Lambda, cutting operational costs by 70%.
• Engineered fully automated ETL pipelines using Python and FastAPI for cross-organizational data analysis.