Hasan Naqvi

• Designed and deployed RAG-based AI chatbot systems using LangChain, FAISS, and OpenAI LLMs, reducing hallucinations and improving contextual response accuracy by 30%. • Built end-to-end pipelines for document ingestion, embedding generation, and semantic retrieval. • Implemented JWT-based authentication & role-based access control using Django. • Integrated speech-to-text pipelines (Whisper) for multimodal interaction. • Reduced response latency by 20% via prompt optimization and caching.

• Processed and annotated 3500+ audio samples, enabling training of NLP models, RASA chat and TensorFlow models handling 60k+ user conversations and improving deployment efficiency using Docker. • Built data validation and QA pipelines of excel files through keyword segmentation, reducing noisy and duplicate data by 15%. • Deployed ML/LLM services using FastAPI with REST endpoints for scalable frontend systems. • Improved chatbot response accuracy to 92% through iterative fine-tuning and error analysis.

• Collaborated as a Machine Learning Trainee developing ML algorithms and Neural networks from scratch, achieving 95% accuracy. • Developed neural networks with16 layers, achieving an accuracy of 90%. • Understand core fundamentals of Machine learning, deep learning and applying models into data analysis. • Worked with computer vision frameworks (opencv, Yolo) on cancer datasets achieving a high parameter accuracy using ensembled learning and regularization with 95% accuracy under 8 epochs.

• Collaborated on Etherea project focusing on building an AI society with 15 agents following industry standards and git version controlling. • Demonstrated technical integration and handling issues virtually over 35+ meetings via teams for regular follow-ups. • Conducted extensive research and utilized Openai LLM to develop agent character scripts and conversations.