Vision Pipelines in Production: Serving and Optimizations - AI in Production

Explore vision pipelines in production with a focus on serving and optimizations in this 13-minute conference talk by Biswaroop Bhattacharjee, Senior ML Engineer at Prem AI. Delve into the transition from solution development to production for vision models, covering topics such as fine-tuning LORAs, upscaling pipelines, constraints-based generations, and step-by-step enhancements. Learn how to achieve optimal performance and quality for a production-ready service. Gain insights from Bhattacharjee's experience in ML platforms, distributed systems, and various AI applications including conversational voice AI, chatbots, and fashion hyperforecasting. Discover practical approaches to improve vision model performance and prepare for real-world deployment challenges.

Intro
Outline
Problem Statement
Upscaling Pipeline
Upscaling Results
Improving Performance
Prem Grand Program
Recap