Beyond Model Accuracy: Overcoming Challenges of AI Inference

Inference is an extremely important part of every AI-enabled product or service. Even the best AI model isn't worth much if it can't meet the latency or bandwidth requirements set out by the business or use case it's serving. In real-life operation, the cost associated with serving (both initial investment and ongoing costs) matter tremendously and can make or break your business case, making all the difference between a successful product and — especially for many startups — bankruptcy. This panel gathers business leaders from across the world to share their experiences in overcoming challenges of AI system deployment. We'll cover a broad range of topics, starting with challenges of MLOps to complexities of model selection and optimization or hosting of countless of AI pipelines.