Reliable AI at Scale: Lessons from Dataiku & Mistral AI

Join us on October 27th at 6:30 pm at our Paris office to explore what it really takes to make AI work in the real world.
Maxime Appé and Thierry Gengler from Dataiku will show how to keep autonomous LLM agents under control without sacrificing their flexibility. Then Harizo Rajaona from Mistral AI will take us behind the scenes of serving a text-to-speech model at scale, one audio frame at a time.

6:30 – 7:15 pm | Maxime Appé\, Technical Product Manager\, and Thierry Gengler\, Engineering Manager\, Dataiku
Taming Rogue LLMs: Closing the Agent Reliability Gap
LLM agents promise flexible, autonomous problem-solving, but that same creativity becomes a liability when enterprise processes demand predictable, reliable behavior. Stochastic drift, endless retry loops, lost context, ignored constraints, latency, overly complex designs: in production, a lot can go wrong.
This talk shows how Dataiku Structured Visual Agents balance generative flexibility with structured control, turning unpredictable agent behavior into transparent, reliable, and maintainable enterprise workflows.

7:15 – 8:00 pm | Harizo Rajaona\, Software Engineer\, Mistral AI
Deploying and Running a TTS Model: The Fun Part
Text-to-speech (TTS) models have improved dramatically in recent years, driven by the rise of Transformer architectures. Training one might sound like the hard part, but wait until you have to serve it at scale!
TTS comes with its own set of challenges: tokenizing audio frames without information loss, combining a decoder with flow-matching transformers, and handling real-time streaming.
This talk walks through these challenges and the engineering choices involved in building and serving Voxtral TTS, proving that making GPUs actually talk is hard but a lot of fun!

The talks will be followed by open discussion and networking. Come ready to ask questions.

8 - 10PM: Meet & Greet with drinks

Dataiku will only use your personal information to provide the product or service you requested and contact you with related content that may interest you. You may unsubscribe from these communications at any time. For more information on unsubscribing and how we protect and respect your privacy, check out our Privacy Policy.