Ansh Manwani@ansh21
1
Events attended
1
Submissions made
Pakistan
1 year of experience
About me
I am a Computer Science undergraduate and AI developer passionate about building intelligent applications powered by Machine Learning, Generative AI, Natural Language Processing, and Computer Vision. My experience includes developing AI-driven solutions such as a Job Application Assistant using LLMs and NLP, a Retrieval-Augmented Generation (RAG) Document Q&A System, a YOLOv8-based Smart Helmet Detection System, and conversational AI chatbots. I work with technologies including Python, TensorFlow, PyTorch, Hugging Face Transformers, LangChain, OpenCV, vector databases, and cloud platforms such as AWS and Google Cloud. I enjoy transforming complex AI concepts into practical products that solve real-world problems. My interests span Generative AI, prompt engineering, computer vision, NLP, MLOps, and scalable AI systems. Through hackathons and collaborative projects, I aim to build innovative solutions that create measurable impact while continuously expanding my expertise in cutting-edge AI technologies. I am excited to connect with builders, researchers, and innovators who share a passion for advancing the future of AI.
🤓 Latest Submissions
AI Ecommerce Customer Support System
AI Ecommerce Customer Support System is an autonomous customer support solution built for the Band of Agents Hackathon 2026, targeting Track One: Enterprise Customer Support. The system deploys three specialised AI agents — Triage Agent, Support Agent, and Manager Agent — that collaborate through Band Platform's shared room infrastructure to handle customer complaints end to end, without any human intervention. The Triage Agent classifies incoming complaints by category and priority. The Support Agent searches an FAQ knowledge base and drafts a resolution with a confidence rating. The Manager Agent evaluates every response through a 100-point quality model — approving it or sending it back with specific revision feedback. This self-correcting loop runs automatically until the quality bar is met. Built using LangGraph, Qwen-2.5 via Featherless AI, and GPT-4o via AI ML API, the system reduces average resolution time from 45 minutes to under 30 seconds, with zero manual steps and 100% quality coverage on every ticket. Band Platform sits at the centre of the entire architecture — providing shared rooms, @mention routing, persistent memory, and a unified audit trail across all three agents.
19 Jun 2026
