The Truth About Long Context & Real-World Evals - with Apoorva Joshi (MongoDB)
Автор: Shipped AI
Загружено: 2025-11-24
Просмотров: 32
In this episode of Shipped AI, we sit down with Apoorva Joshi, Senior AI Developer Advocate at MongoDB, to uncover the real constraints behind building reliable AI systems.
We examine why long-context models fail earlier than expected, how hybrid retrieval architectures keep systems stable, and why benchmarks rarely reflect real-world performance. Apoorva shares practical lessons on evals, observability, and designing agentic workflows that remain dependable in production.
🔹 Topics covered:
• The truth about “1M token” long-context claims
• Hybrid retrieval and why it’s now the default
• Benchmarks vs. evals for real-world reliability
• Designing agentic flows without unnecessary complexity
🎙 Hosted by Productship — helping teams make AI products people want, and ship them reliably.
#AI #LLM #RAG #Evals #MongoDB
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке:
