MLCon NYC 2025: From Model to Market

Speaking at MLCon NYC 2025 was an opportunity to share the practical engineering challenges of taking a generative AI concept and deploying it as a scalable SaaS product.

Deploying at Scale

My session, "Deploying AI-Powered Video & Audio Models: A SaaS Perspective," focused on the orchestration and cloud scaling required for intensive gen-AI workloads.

Key Deployment Strategies:

• Cloud Scaling: Managing GPU-intensive workloads across distributed environments.
• Model Orchestration: Synchronizing video and audio generation models into a unified service package.
• Reproduction: Ensuring masterable, reproducible workflows for enterprise-grade models.

The SaaS Perspective

Building real-world AI systems requires moving beyond the initial research model. We discussed how to handle:

• Service Packaging: Optimizing FastAPI backends for high-concurrency gen-AI tasks.
• Safety & Security: Hardening models for enterprise use cases.

Representing Pattern at MLCon was a great chance to connect with the ML Builders community and discuss the future of Agentic Engineering.