Stenobird

Episode

A Playground for AI/ML Engineers

Podcast
MLOps.community
Published
Jan 23, 2026
Duration seconds
3281
Processing state
processed
Canonical source
https://podcasters.spotify.com/pod/show/mlops/episodes/A-Playground-for-AIML-Engineers-e3e2qam
Audio
https://anchor.fm/s/174cb1b8/podcast/play/114435862/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-0-23%2F416681357-44100-2-4e1efed8e8014.mp3
JSON
/v1/public/podcasts/mlops-community/episodes/a-playground-for-ai-ml-engineers
Markdown
/podcast/mlops-community/a-playground-for-ai-ml-engineers.md

Actions

  • POST https://stenobird.com/v1/public/podcasts/mlops-community/episodes/a-playground-for-ai-ml-engineers/transcription-requests
    Idempotently request low-priority transcript generation for this episode.
  • GET https://stenobird.com/podcast/mlops-community/a-playground-for-ai-ml-engineers.md
    Read the agent-friendly Markdown representation of this episode resource.

Summary

Learn how Hotmart leverages LLMs and classic NLP to transform passive digital courses into interactive AI-driven learning experiences. This discussion explores the transition from 'AI as a feature' to 'Agent as a product' within a large-scale creator ecosystem.

Topics

  • Generative AI
  • AI Agents
  • MLOps
  • NLP
  • RAG
  • Context Engineering
  • EdTech
  • Product Strategy

Highlights

  • Main idea: Moving from passive content consumption to active learning through AI tutors that use specific course knowledge bases
  • Practical takeaway: Use 'context engineering' to inject real-time user data and vector database insights into agent prompts for fluid interactions
  • Failure mode: Relying solely on LLMs for all tasks; hybrid approaches using classic NLP (like spaCy) remain essential for specific, cost-effective production needs
  • Main idea: The 'Agent as a Product' model allows creators to monetize specialized AI agents built on their unique instructional content
  • Practical takeaway: Building AI agents requires a flexible infrastructure capable of swapping endpoints and integrating tool-augmented capabilities

Chapters

  1. 1:00 Hotmart Data Science Challenges: An overview of using machine learning for fraud detection, content moderation, and recommendation systems in the education industry.
  2. 5:10 LLMs vs spaCy: Discussing the trade-offs between powerful LLMs and efficient, classic NLP frameworks for specific production tasks.
  3. 9:15 Use Cases in Production: How entity extraction and model insights are routed to dashboards to drive business value.
  4. 13:10 The AI Tutor Experience: Implementing RAG-based agents that use specific course content to provide accurate, hallucination-free tutoring.
  5. 17:10 Edge Cases in AI Products: The difficulty of managing edge cases when providing AI-driven insights to content creators.
  6. 21:30 Driving Student Retention: How interactive AI interfaces transform passive video watching into an engaging, critical-thinking exercise.
  7. 30:05 The Shift to Agentic Products: Preparing for a rapid industry shift where developers must be ready to rebuild architectures around agentic workflows.
  8. 34:05 Tool-Augmented Agent Approach: Designing modular infrastructure where LLMs act as components that can call specific tools and endpoints.