Shahzaib Saqib W.

I build robust and reliable Agentic AI and Computer Vision systems that run in the real world, not just in demos. My vision systems run on Fortune 500 factory floors, and my research on LLM reliability has been published at ICLR 2026 (Oral) and COLM 2026. I've spent 7+ years shipping AI at companies like Paramount and for manufacturers like Honda, Carrier, and Schneider Electric. I hold an M.S. in Applied Data Science from University of Southern California as a Fulbright Scholar. Agentic AI: + Autonomous agents and multi-step workflows that plan, use tools, and take action + LLM apps, RAG pipelines, and fine-tuned models built for your data + Evaluation and guardrails that make agents reliable, cutting hallucinations and failures before they reach users Computer Vision: + Object detection, activity recognition, and pose estimation + Real-time video analytics and quality inspection + Edge deployment optimized for speed and cost Proven results: + Built Assembly Copilot, a real-time vision system with 98% detection accuracy, deployed for Honda, Carrier, and Schneider Electric + Cut manual data annotation by 80% with synthetic data generation + Reduced edge inference latency by 66% with under 1.5% accuracy loss + At Paramount, improved content performance forecasting by up to 20% and built GenAI explainability tools Stack: Python, PyTorch, Hugging Face, vLLM, TensorRT, AWS SageMaker, GCP Vertex AI As a Sr. Solutions Architect of an AI custom solutions development agency, I design production AI for Fortune 500 clients, so I scope projects around business outcomes, not just code. Expect clear communication, fast iteration, and systems built to scale. Have an agent to build or a vision problem to solve? Message me and let's get started!