Swayam Chidrawar

• Built scalable evaluation pipeline using AWS ECS Fargate, SQS, and EFS with automated visualization dashboards, enabling rapid cost-performance optimization without manual testing or customer input.
• Architected autonomous RAG optimization framework that auto-generates QA pairs, performs intelligent clustering, and evaluates recall@k metrics to determine optimal model and configuration settings for any customer dataset.
• Spearheaded multiple Proof of Concept (POC) Retrieval-Augmented Generation (RAG) solutions for municipal clients, including San Luis Obispo and San Diego, addressing specific city challenges utilizing graph based RAG.
• Trained autoencoder and regression models using scikit-learn, TensorFlow, and SageMaker to assess property values for the Alameda County Assessor’s Office
• Built an end-to-end pipeline for extracting data from podcasts and videos into vector databases for RAG applications using AWS Transcribe, Textract and OpenCV, facilitating advanced content retrieval
• Developed MyECISO, a chatbot that evaluates and provides feedback to clients’ cybersecurity teams, based on the NIST cybersecurity framework, in collaboration with San Diego Cyber Center of Excellence
• Apache Beam

• Fine-tuned LLMs using PyTorch and Transformers, improving text generation quality and classification accuracy
• Developed and optimized large-scale synthetic data generation pipelines for training LLMs, enhancing training efficiency and reducing data preparation time
• Developed a custom scoring system combining lexical and semantic search scores, improving search accuracy for Graph RAG

• Fine-tuned LLMs using PyTorch and Transformers, improving text generation quality and classification accuracy
• Developed and optimized large-scale synthetic data generation pipelines for training LLMs, enhancing training efficiency and reducing data preparation time
• Developed a custom scoring system combining lexical and semantic search scores, improving search accuracy for Graph RAG

GPA: 3.8