Field Report: Putting AI Pipelines into Production

How do you take an AI pipeline from prototype to something you can trust in production? Daniel Altunay, AI and ML engineer and founder of FilterGenie, walks through how he designs, evaluates and monitors pipelines that use LLMs and vision-language models to pull structured data out of documents. He covers choosing methods and models, benchmarking, trading quality against latency and cost, validating outputs and handling errors, then monitoring and catching regressions once the pipeline is live. Hosted by datacraft, the Paris data science community. Monday 16 November, 18:00 to 20:00. In French. Free; registration is approval-gated and the exact Paris address is shared once you're accepted. Why go: a practitioner's view of document extraction with LLMs from someone running it in production. Best for AI engineers trying to get a document pipeline out of the notebook.