Chittal Karuppiah

- Built an AI-based tutoring assistant using Amazon Bedrock Agents for real-time chat, notes, quiz, and flashcard generation from YouTube video transcripts. - Utilized Langsmith to monitor and track agent behavior, enhancing observability and enabling detailed analysis. - Utilized Elevenlabs to integrate speech-to-text on the chat window. - Participated in weekly advanced AI sessions covering LLMs, agent design, cloud deployment, and AI startup case studies.

- Developed Synthesize Chatbot, a multi-agent LLM system that merges and contrasts outputs from multiple LLMs (e.g., GPT-4, Claude, LLaMA) into a single, accurate response, improving relevance and user trust over single-model outputs. - Developed an agentic AI low-code/no-code platform leveraging PydanticAI to define and create customizable conversational AI agents, enabling users to create agents with the ability to add knowledge sources via web page links or document uploads for richer context. - Built a RAG-based chatbot using Pinecone vector DB, OpenAI embeddings, and FastAPI, enabling real-time knowledge retrieval and personalized responses from custom document sets. - Containerized services as reusable Python packages and Dockerized microservices, improving modularity and collaboration across teams. - Deployed applications using AWS ECR, ECS Fargate, Application Load Balancer (ALB), and CI/CD pipelines.

- Built scalable backend systems using Flask and Django REST Framework for flight booking and assistance platforms - Integrated PostgreSQL, MySQL, and MongoDB with ORM for robust data handling - Designed async workflows with Google Cloud Tasks, Pub/Sub, Celery, and SNS - Deployed microservices via Docker on GCP, using App Engine for scalability - Improved API performance by 80% with Elasticsearch and Couchbase - Worked in Agile teams with GitHub, CI/CD, and cross-functional collaboration (React.js, QA)

I have gained hands-on experience in developing prediction and classification projects and building chatbots using LLM and RAG.