● Implemented a LLM-based system that adapts to user feedback, ensuring generated reports align with the user's unique style. Migrated the system from monolithic to microservices architecture using ECS, reducing system load by 40%. This migration streamlined independent service updates and improved scalability● Developed an engaging frontend, increasing user interaction by 35%, and optimized an ETL pipeline, improving data transfer efficiency by 80% and enabling data analysis for the sales team using Supabase and Metabase● Designed a rag based pipeline that generates personalized, demographically tailored medical reports, enhancing report accuracy and relevance based on patient history and contributing to a $1M ARR milestone● Created a company-wide protocol leveraging prompt engineering and function calling with LLM APIs, streamlining and democratizing the testing process by enabling non-engineers to actively participate, boosting productivity by 50%● Architected a modular backend structure to support high traffic loads, leveraging microservices to allow independent scalability, cutting infrastructure costs by 25%.

● Engineered and fine-tuned LLMs such as phi-3, openELM, and Llama3 using torch and transformers, implementing QLoRA and DPO (Direct Preference Optimization) techniques to significantly enhance the quality and automation of medical report generation● Deployed a REST API-based medical report automation on Azure, optimizing API endpoints to improve processing efficiency by 15%. Utilized caching and load-balancing for high traffic, ensuring scalable and fault-tolerant performance with efficient data structure management● Developed a change-tracking tool in Python using difflib and fuzzywuzzy libraries, wrapped in a FastAPI framework for API-driven interactions. This tool implemented customized data structures for tracking and comparing revisions, achieving $50,000 in annual cost savings through reduced manual intervention● Engineered a REST API for medical term translations using MySQL and SQLAlchemy ORM, employing indexing and optimized query patterns to reduce response times by 30%. Ensured robust data integrity and fault tolerance to handle high concurrency● Utilized Azure Functions, Blob Storage, and Cognitive Services to create a scalable, efficient workflow pipeline, significantly improving data processing throughput and reducing operational costs by 20%● Automated infrastructure deployment using Terraform for a consistent, reproducible architecture on Azure, focusing on scalability, fault tolerance, and maintaining low-latency responses

● Architected an LMS similar to Canvas, featuring a discussion board inspired by Stack Overflow. Applied object-oriented principles to boost modularity, and user engagement rose by 45% due to a well-structured user interface backed by a scalable API design.● Built a robust REST API using Flask and MySQL for LMS features, including complex notification services with optimized message-passing algorithms to support large-scale push notifications. Ensured data consistency and transaction management for high security and scalability.● Integrated JSON for seamless data exchange and leveraged Socket.io for real-time, low-latency communication, ensuring responsive user feedback and robust system interaction under high-load conditions.