Deepansh Jha

Energetic and innovative Generative AI Engineer with a strong background in AI, specializing in vector databases, LLM frameworks, and cloud deployments. Eager to apply technical skills and creativity to develop state-of-the-art AI solutions within a dynamic team setting.

• Innovative AI Solutions: Successfully developed and deployed cutting-edge AI applications, demonstrating a strong impact in enhancing user engagement and operational efficiency
• Technical Leadership: Led teams and projects to pioneer the use of generative AI technologies in practical, real-world applications

Managed and optimized vector databases such as ChromaDB, Pinecone, and AstraDB, enhancing AI applications' data retrieval capabilities. Handled DataStax Cassandra DB in production environments, ensuring optimal performance and reliability for AI-driven applications. Developed scalable AI solutions using LangChain and LlamaIndex frameworks, demonstrating expertise in generative AI technologies. Implemented and fine-tuned both open-source and paid LLM models, customizing solutions to meet specific project requirements and performance goals. Leveraged AWS EC2 for deploying AI models, utilizing cloud services to ensure scalability and reliability of AI applications.

RAG Model Project with DataStax Databases and Vector Embedding:

• Overview: Developed a cutting-edge Retrieval-Augmented Generation (RAG) model integrating advanced AI tools and frameworks.
  Technologies: Astra DB, Ollama, Hugging Face, LangChain, custom fine-tuned PHI3 LLM model using the Open Assistant dataset.
  Outcome: Achieved a 50% improvement in Enhanced skills and knowledge in AI and LLM deployment.
  

Web Application for a Conversational AI system using LLM and DataStax Databases and Vector Embedding:

• Overview: Developed a Astrology conversational AI web app using Flask, integrating advanced AI and NLP tools for context-aware responses.
• Technologies: Flask, Python, AstraDB, Cassandra, Hugging Face Inference API, Langchain Embedding, LlamaIndex, AWS, dotenv
• Outcome: Improved response accuracy by 40% and reduced query processing time by 50%, enhancing user satisfaction and application efficiency.
  

Conversational Medical AI Chatbot with Pinecone and Langchain:

• Overview: Developed a web-based conversational AI application using Flask, integrating vector search and retrieval-augmented generation to deliver accurate and context-aware responses.
• Technologies: Flask, Python, Pinecone, dotenv, Hugging Face embeddings, CTransformers, Langchain
• Outcome: Improved response accuracy by 45% and reduced query processing time by 55%, significantly enhancing user experience and application performance.

Unreal Engine 5 Plugin for Conversational AI using Custom LLM Model AND PDFs:

• Overview: Developed a third-person game in Unreal Engine 5 featuring a Retrieval-Augmented Generation (RAG) chat system. Players can interact with a custom LLM model integrated as a plugin. Upon reaching a specific platform, an automatic UI opens, allowing local chat with the LLM without using any external API.
• Technologies: Unreal Engine 5, Open Source LLM Model (HuggingFace), Texture, Material, Blueprint Class, Blueprint Variables, Custom Widget, Event Graph, Nodes, Python Script, Langchain, PineconeDB
• Outcome: Improved in-game interaction efficiency by 35% and enhanced user satisfaction by 45% through seamless, local Retrieval Augmented Generation integration.

• iNeuron (Generative AI Foundation)
• IIITB & UpGrad (Data Science Programming Bootcamp)
• IIITB & UpGrad (Data Toolkit)
• IIITB & UpGrad (Data Engineering-I)

• Singing
• Snooker
• Photography
• Content Creator