Sabih Ali

AI Engineer with hands-on experience shipping agentic RAG systems and GPU-accelerated computer vision pipelines end-to-end, from model training through production deployment. Builds multi-LLM orchestration pipelines (LangGraph, hybrid retrieval, Reciprocal Rank Fusion) alongside real-time CV systems (YOLO, multi-object tracking, TensorRT/ONNX inference), and guardrailed, CI-evaluated LLM agents, deployed into microservices-based FastAPI backends. Currently extracting spatiotemporal features from live match footage to power agentic tactical-analysis tooling, and has previously optimized inference pipelines to a measured ~60% latency reduction and ~1.8x speedup on production hardware.

Junior AI Engineer at Rapids AI, SINES, NUST (2026-06 – Present)

• Built Pitchside, an agentic RAG football chatbot with a multi-LLM orchestrator-generator-judge pipeline to minimize hallucination.
• Built RapidEye, a real-time multi-camera video analytics platform running 7 TensorRT-optimized models on one GPU.
• Built AutoPulse, an agentic AI layer in a live dealership CRM automating lead replies, qualification and booking.
• Delivered all three projects within a microservices architecture spanning Python backends and modern web frontends.

AI Intern at EmbedAIoT, SINES, NUST (2025-06 – 2025-09)

• Implemented small-object detection pipelines using YOLO-based architecture for surveillance imagery.
• Optimized training using transfer learning and augmentation strategies to improve recall on sub-32px objects, achieving an 83.4% mAP50 score.
• Contributed to early-stage reinforcement learning control research for autonomous rocket landing.

Bachelor of Engineering in Software Engineering – National University of Sciences and Technology - Islamabad