Manya Jain

Engineered production AI microservices for legal clause detection and multimodal document classification using transformer-based foundation models, computer-vision pipelines, Python/C++, and REST APIs. Improved multimodal model inference throughput and evaluation performance via GPU-accelerated PyTorch pipelines, model fine-tuning, multiprocessing I/O, caching, and distributed microservices, achieving a 40% throughput increase and 35% latency reduction across 200+ documents. Developed generative AI workflows for document understanding and information retrieval using LLMs, semantic search, and tool-enabled retrieval pipelines for scalable enterprise deployment.

Designed and implemented enterprise document-processing automation using shell scripting and CI/CD pipelines to enable continuous integration and deployment, improving throughput by 60% and reducing analysis time by 50%. Translated requirements into technical specifications for service integration using scalable REST APIs and regression testing checks, improving model accuracy by 20% and reducing integration time by 35%.

Executed data-driven research to improve evaluation rubric quality for predictive modeling initiatives. Collected and processed large-scale datasets using Python and Selenium, then analyzed user acceptance testing results from 45 participants to define and refine grading criteria, improving evaluation accuracy informed by participant feedback.