William – Python, LLM, LangGraph

It's AI-driven automation software for U.S. healthcare providers, designed for eligibility verification, benefits analysis, and administrative workflows. The system integrates LLMs and external APIs to process unstructured and structured data, automate repetitive tasks, and generate actionable insights.

• Designed and implemented automation bots for healthcare eligibility verification and benefits workflows.
• Integrated OpenAI APIs to process unstructured medical and insurance data.
• Built data processing pipelines with Pandas to handle large-scale healthcare datasets.
• Developed back-end services to orchestrate AI-driven task execution.
• Deployed and maintained automation solutions on AWS infrastructure.
• Integrated bots with third-party healthcare systems and internal platforms.
• Improved operational efficiency by reducing manual processing time through automation.
• Monitored system reliability and optimized workflows for production stability.

It's an enterprise AI platform for large-scale text and image processing, model orchestration, and observability of LLM-based services.

• Designed and implemented unified LLM integration layers using LiteLLM and Langfuse.
• Maintained observability pipelines to monitor prompts, responses, latency, and costs.
• Integrated multiple language models for text and image processing workflows.
• Extended database schemas to store and analyze AI request metadata.
• Developed high-throughput backend services with FastAPI for model orchestration.
• Implemented monitoring and alerting mechanisms to ensure AI service reliability.
• Enhanced system scalability and cost efficiency through usage analytics and optimization.

It's a large-scale data collection and validation platform for training and evaluating LLMs. The system automates web data ingestion, cleaning, and normalization from multiple sources, ensuring high-quality datasets for model training.

• Led a data engineering team to deliver large-scale, high-quality datasets for LLM training.
• Designed and implemented automated web scraping pipelines for multi-source data collection.
• Built data cleaning and normalization workflows to ensure consistency and reliability.
• Developed AI agents to monitor dataset quality and validate ingestion metrics.
• Created FastAPI services for dataset management and reporting.
• Established internal quality standards and validation processes for training data.
• Improved data accuracy and reduced noise through automated verification pipelines.
• Collaborated cross-functionally with engineering and AI research teams to align objectives and delivery.

It's an AI and automation platform for large-scale mobile device testing and quality validation. The system collects and processes test execution data, generates training datasets, and delivers predictive models to optimize test coverage and defect detection.

• Built ML models to analyze test data and optimize defect detection.
• Created and maintained large-scale datasets for model training and validation.
• Developed REST APIs with Flask and FastAPI to serve ML models in production.
• Implemented CI/CD pipelines for containerized applications using Docker.
• Integrated ML services with internal Motorola testing platforms.
• Monitored model performance and improved accuracy through iterative fine-tuning.
• Designed and managed the complete lifecycle for trained and deployed LLMs.

It's an AI-powered document analysis and question-answering platform for extracting and understanding information from images and PDF files.

• Designed and fine-tuned multimodal LLMs for document and image understanding.
• Built data ingestion pipelines to process PDFs, images, and structured data inputs.
• Developed REST APIs with FastAPI for real-time inference and question answering.
• Applied few-shot learning techniques to improve model response accuracy.
• Trained CNN models using transfer learning for visual feature extraction.
• Deployed and managed AI workloads on Azure virtual machines.
• Integrated SQL databases to store user feedback and training data.
• Enhanced system performance through continuous optimization and model retraining.