Yash Pal

Building Conversational AI solutions for healthcare that automate patient interactions, streamline appointment workflows, and replace manual receptionist tasks for clinics and hospitals. Engineered voice and chat-based AI receptionists capable of performing real actions like booking, rescheduling, and canceling appointments, answering FAQs, and handling multi-turn dialogues with human-like empathy and contextual memory. Designed and deployed multi-agent architectures with RAG, dynamic context extraction, and prompt chaining, improving conversation accuracy and intent fulfillment by 35%. Integrated Twilio for telephony, Vocode for real-time voice synthesis, and FastAPI microservices for action orchestration and backend communication. Optimized inference and latency using Redis caching and MongoDB-based session tracking, enabling seamless real-time patient engagement. Leveraged Azure cloud, LLMs (GPT family), and custom prompt engineering for scalable, secure, and HIPAA-aligned deployment of AI reception systems.

Contributed to cutting-edge Generative AI, Computer Vision, and Intelligent Automation initiatives, delivering scalable and production-ready AI solutions that enhanced client productivity and operational efficiency. Led development of Chat-with-your-Data systems enabling users to query and extract insights from structured (SQL, CSV) and unstructured (PDF, DOCX, Web) data using RAG pipelines, custom LLM prompts, and embedding-based retrieval, improving data accessibility and decision-making speed by 40%. Designed and deployed Lip-Sync AI and MuseTalk-inspired facial animation models using GANs, attention mechanisms, and image-to-image translation, enhancing synchronization accuracy by 25% for media automation use cases. Built autonomous multi-agent workflows with LangGraph and CrewAI to automate repetitive business tasks, simulate human decision-making, and enable context-aware process flows, cutting manual intervention time by 30%.

Contributed to the design and deployment of an end-to-end AI platform powering computer vision, NLP, and enterprise automation use cases across multiple industries. Delivered scalable, production-ready models and MLOps frameworks that accelerated AI adoption and operational efficiency. Architected the backend of a low-code/no-code AI platform using Argo Workflows, enabling seamless ML pipeline orchestration and automated task execution, reducing model deployment time by 40%. Integrated MLflow for experiment tracking, model versioning, and centralized registry, improving collaboration and model reproducibility across teams. Processed high-volume image sensor datasets, converting raw visual inputs into structured feature vectors optimized for downstream analytics and predictive modeling. Built EDE (Electrical Diagram Extractor) leveraging OpenCV, custom CNN architectures, and OCR pipelines to automatically interpret and extract entities from electrical diagrams, reducing manual analysis effort by 60%. Developed Retrieval-Augmented Generation (RAG) prototypes integrating LLMs for contextual document retrieval and knowledge-grounded response generation in enterprise search scenarios. Deployed a transformer-based sentiment classifier (BERT) to categorize customer feedback with 92% accuracy, enhancing actionable insight generation. Built SKU demand forecasting models using ARIMA, ETS, and hybrid approaches, improving inventory planning precision by 25% for retail and logistics operations.

Developed AI-powered computer vision solutions for e-commerce automation, enhancing product imagery, detection accuracy, and content onboarding efficiency. Built and optimized deep learning models for object detection, segmentation, and classification, improving recognition accuracy by 28% and reducing manual curation time by 45%. Implemented GANs and Diffusion-based architectures to generate high-quality, realistic product visuals, accelerating catalog creation for online retailers. Managed end-to-end development of Computer Vision pipelines using Python, OpenCV, PyTorch, and TensorFlow, integrating them into scalable REST API frameworks.