Building and testing LLM based agents to facilitate medical information retrieval.

* Developed a SvelteKit website for educating beginners in programming, building the entire backend with C++ and Poco libraries to manage WebSocket connections for user management and enable code execution in Python, R, and Rust directly on the site.

* Engineered microservices in Go, including a REST API for Stripe payment processing, a Discord bot leveraging the Discord API for subscription authentication, and an interactive coding platform with dynamic project generation akin to LeetCode.

* Deployed backend microservices on DigitalOcean's cloud platform, configured Nginx as a reverse proxy for directing web traffic, and orchestrated containerized services with Docker Compose for scalable, efficient management.

* Automated fine-tuning of OpenAI's GPT-3.5-turbo model to classify large-scale unstructured Ad data, built a pipeline to convert any input to JSONL, and utilized RestAPI with multi-threading for fine-tuning requests, achieving a 92% accuracy rate.

* Deployed the product on an AWS EC2 container as a SaaS, using Docker Compose for multi-container orchestration and Apache Airflow to enable one-click execution from the front-end web page.

* Designed a product interface webpage adaptable to both desktop and mobile platforms using React and MaterialUI, incorporating Redux for state management and Hasura GraphQL to interface with the PostgreSQL database for data storage and retrieval.

* Wrote a high-performance recommendation engine using XGBoost, optimizing for AUC and conversion rate; Implemented rigorous feature engineering, preprocessing, and ensembling to enhance model accuracy across large datasets.

* Led a significant ETL refactor, achieving a 65% improvement in processing time by utilizing REST APIs to extract clinic data from multiple hospitals, transforming and normalizing datasets with Python (NumPy, Pandas), and loading them into a centralized MongoDB database.

* Engaged in daily Scrum stand-ups and sprint planning sessions, applying Agile principles and practices such as Test-Driven Development and continuous integration to deliver error-free code on the first try, reducing QA cycle time by 30% and improving overall sprint velocity.

* Developed a Language-Agnostic Search Retrieval algorithm using tokenization, word embeddings, and custom-trained transformers to capture cross-language semantics, along with zero-shot learning and human-in-the-loop to ensure accurate document retrieval.

* Adapted the system to run on SLURM compute infrastructure using nvidia-smi and parallelization, leveraging HPC for scalability, scripting for automation, performance tuning for GPU optimization, and implementing distributed computing to handle large-scale processing.