Harsh Singh

Amazon GenAI Recommender System: Engineered a hybrid recommendation engine combining collaborative filtering, content-based filtering (using semantic embeddings), and LLM re-ranking via Groq API to enhance product relevance. Implemented FAISS for high-speed similarity search and integrated Sentence Transformers to enable semantic search capabilities, significantly improving cross-selling efficiency. Developed a scalable, modular architecture with caching mechanisms to optimize performance for large datasets and cold-start scenarios. Deployed an interactive Streamlit interface providing real-time recommendations, advanced filtering, and a product analytics dashboard for category and price analysis.

Agentic RAG System: Architected an Agentic RAG system utilizing LlamaIndex, Gemini, and Hugging Face embeddings to optimize document retrieval and generation. Designed smart query handling logic to dynamically route requests between retrieval and summarization, reducing hallucinations and improving response latency. Enhanced system adaptability for complex queries, ensuring precise context integration and efficient task coordination.

Developed a RAG-based healthcare chatbot and a multi-agent system using CrewAI for efficient task coordination. Contributed to AI projects emphasizing scalability, user-centric design, and impactful LLM applications.