Imran Nadeem

AI/ML Engineer building agentic AI systems and multimodal models that fuse text, voice, and vision into a single intelligent pipeline. Experienced across the full AI development lifecycle — data preprocessing, EDA, feature engineering, model fine-tuning, RAG pipeline design, and deploying LLM-powered agents via FastAPI and Docker.

At Funsol

Technologies, built Mentimotive, a real-time multimodal emotion-AI system, as part of an industry-academia capstone; at Genesys Research Lab, built a real-time AI fitness coach using computer vision and pose estimation. Also built an AI Medical Chatbot using Retrieval-Augmented Generation (RAG) to minimize hallucination in a sensitive domain. Focused on shipping production-grade agentic and multimodal AI systems end to end.

AI Developer - Funsol Technologies Pvt. Ltd.

(2025-10 - 2026-05)

Industry–Academia Capstone / FYP

• Built Mentimotive, a multimodal Mental Health & Motivational AI Companion fusing text, voice, and video emotion signals (DistilBERT, Wav2Vec 2.0, Whisper) into a single real-time inference pipeline.
• Architected an agentic FastAPI backend with async endpoints, JWT auth, and modular services for sentiment analysis, personalised response generation, and live interaction.
• Implemented a context-aware conversational memory system with LangChain + ChromaDB, improving multi-turn response relevance and continuity.
• Deployed the full multimodal pipeline via Docker for reproducible builds; collaborated in a 4-engineer agile team under industry-academia standards.

AI Fitness Coach - Genesys Research Lab

Computer Vision Project

• Built a real-time AI exercise analysis system using MediaPipe Pose Estimation and OpenCV for human pose tracking, rep counting, and form correction.
• Developed a configuration-driven exercise engine supporting multiple workouts with dynamic joint-angle calculations, ROM analysis, and movement validation.
• Implemented DTW-based movement comparison against reference videos and generated quality scores using movement similarity, range of motion, and form accuracy.
• Designed real-time feedback detection for incorrect posture including elbow flaring, body leaning, and improper exercise depth using biomechanical rules.
• Built scalable backend infrastructure using FastAPI, WebSockets, and PostgreSQL for live session management and performance tracking.

BSc - Computer Science - National University of Technology (NUTECH) (2026-06)