Nirmata

At Nirmata, we develop AI-driven multi-agent systems (MoA) for healthcare, education, and e-commerce, integrating deep learning, LLMs, and RAG for intelligent decision-making. ### Current Work: - Multi-Agent AI System: Inspired by Manus AI, we designed an MoA-based architecture, where specialized agents work in parallel, and an Aggregator Agent ensures structured, bias-free outputs. - Medical AI for Skin and Health Diagnostics: Using YOLO and U-Net for skin, hair, and body care analysis, detecting acne, scars, blackheads, and more. Integrated Meditron (LLM for medical Q&A) and a fine-tuned LLaMA model with CoT reasoning for intelligent routing. - Real-Time Heart Monitoring AI: Built deep learning models for ECG arrhythmia detection using ProtoCentral MAX30003 sensors, continuously improving architectures in PyTorch. - AI-Driven E-commerce Aggregation: Developing a D2C platform that gathers product insights from multiple e-commerce platforms for personalized recommendations. - Cloud and Backend Infrastructure: Deployed AI models on AWS (EC2, SageMaker, Lambda), built APIs with FastAPI and Django, and ensured scalability with Docker and Kubernetes. ### Past Projects: - Harmonia – AI Companion: Built a multi-agent AI assistant for daily scheduling and productivity tracking, integrating Google Calendar. - Autonomous Torpedo Navigation: Developed real-time object detection and segmentation with YOLOv5 and U-Net, optimized for NVIDIA Jetson (TensorRT and ONNX). - EduReportAI – AI-Driven Learning Insights: Created a multi-agent system using LangChain, RAG, and Pinecone for personalized education reports. - SuccessSeeker – AI-Powered Interview Coach: Built a hierarchical multi-agent system providing interview practice, feedback, and tracking with GPT-4.5, LangGraph, and AWS. At Nirmata, we are committed to pushing AI innovation for health, learning, and intelligent automation.