LLM Projects

Scalable RAG with GKE and LlamaIndex

Implemented a Q&A Pipeline using the LlamaIndex framework, Qdrant as a vector database, and deployment on Google Kubernetes Engine using a FastAPI app and Dockerfile. Python files from GitHub repositories are loaded into the vector database, and the FastAPI app processes requests within an interactive Streamlit app

AWS RAG Qdrant

Developed a Serverless Application with AWS Lambda and Qdrant for Semantic Search. Utilized AWS API Gateway for API development, Docker for containerization, and created additionally an interactive Streamlit app

CrewAI RAG LangChain Qdrant

Implemented a Retrieval-Augmented Generation (RAG) project for Question-Answering (QA) retrieval, using LangChain and Qdrant as vector store, together with a Research and Writer agent from CrewAI to analyze and summarize research papers

Fine-Tuning Gemma 2B

Fine tuning of Gemma 2B model using quantization and LoRA Adapters (QLoRA) and hosting the model/adapters in a private repo in Hugging Face

Agentic RAG LangChain Pinecone

Implemented a Retrieval-Augmented Generation (RAG) project for Question-Answering (QA) retrieval, using LangChain and Pinecone as vector store. Explored the concept of multi-tenancy and multi-agents workflows including memory

RAG Llama Index

Implemented a Retrieval-Augmented Generation (RAG) project for Question-Answering (QA) retrieval, utilizing Llama Index and Deep Lake vector database

RAG LangChain Ragas

Implemented a RAG Q&A Pipeline project, using LangChain as Framework, FAISS vector database and evaluating the performance of the RAG model using Ragas metrics such as Faithfulness, Answer Relevancy, Context Precision, Context Recall, and Answer Correctness

Q&A and Summarization

Developed a LLM project for Question-Answering and Summarization, employing Whisper ASR (Automatic Speech Recognition) system and Langchain. Implemented an app using Streamlit for local deployment, facilitating the extraction of audio and text information

RAG App with AWS CDK, Qdrant and LlamaIndex

Implemented a RAG Pipeline using AWS CDK as IaC, LlamaIndex framework, Qdrant as a vector database, and deployment on AWS using a FastAPI app and Dockerfile

Agentic RAG Using Claude, LlamaIndex, and Milvus

Implemented an agentic RAG pipeline for Question-Answering retrieval, using LlamaIndex and Milvus as vector store, together with a critic reflection agent

Azure RAG Qdrant

Developed a Serverless Application with LangChain framework, Qdrant as vector database and Azure Function with HttpTrigger

AWS RAG with OpenSearch, Bedrock and Langchain

Implemented an application using Terraform IaC, LangChain as orchestration framework, Bedrock LLM and embedding model and OpenSearch as vector database and endpoint

RAG with Milvus, LlamaIndex and PII Modules

Developed a RAG pipeline using LlamaIndex, Milvus as vector store, and different PII (Personally Identifiable Information) modules from LlamaIndex and Presidio

Multimodal Bill Scan System with AWS Services

Developed a multimodal application using AWS CDK as IaC, Claude 3 Sonnet as multimodal model, DynamoDB as storage. and SQS/SNS for messaging and notifications.