AgentT30 - Overview

Hi there, Welcome to my GitHub 🙋‍♂️

Led and architected scalable Document OCR, Judgment Analytics and Document Prompt modules, serving as the primary technical owner.
Built, prompt engineered and deployed production-grade NLP pipelines using OpenAI GPT, Anthropic Claude, and AWS services (S3, Lambda, Bedrock, SageMaker), achieving 95% extraction accuracy.
Prompt engineered various LLMs and iterated prompts for extracting structured information from OCRd́ legal documents to increase extraction accuracy by 80% by reducing bias and hallucinations.

Engineered and deployed a cost-efficient OCR (Optical Character Recognition) framework for court documents, processing over 250 million documents while reducing operational costs from millions to thousands of dollars.
Finetuned Models like LegalBERT and DeepSeek-R1, resulting in a 15–20% boost in document classification.
Mentored a team of junior engineers and established code review protocols.

Led development of Unicourt’s analytics platform in collaboration with a US-based team, delivering all milestones on schedule.
Integrated large scale datasets from multiple sources using Pandas and Polars to handle large amount of data using minimal resource utilization cutting data processing times
Architected and implemented RESTful APIs integrating the CRM application with various third party applications.
Contributed to an agile software development team, collaborating with developers and testers to achieve project goals.

Developed and added additional functionality on the open source CRM application
Worked closely with US clients to understand CRM application enhancement needs and understood all aspects of business system developments and improvements, including the requirements analysis, design, implementation, testing, and deployment.
Led end-to-end development cycles from requirements gathering to deployment, completing numerous major features during internship period.