Full Workshop: Build Your Own Deep Research Agents
summary
Claims a working end-to-end system: an MCP-powered deep research agent that uses Gemini for grounded web search and YouTube transcription, Firecrawl/Apify for scraping, GitHub content pulls, and tools that write a research.md which a separate deterministic writer workflow turns into polished posts. The pipeline runs locally via an MCP server, includes a compile research tool and an evaluator-optimizer pattern (writer + reviewer loops), logs traces in Opik observability, and reports an LLM judge F1 (tiny evals show perfect scores but need larger datasets).
AI-written summary · about
chapters (21)
- 0:00Introduction and problem space (LinkedIn content and AI slop)
- 3:39Workshop overview and goals
- 5:01Speaker introductions
- 6:09AI engineering problem space and constraints
- 30:14Tech stack and tools (Scraping, Gemini grounding, YouTube/GitHub processing)
- 33:32MCP-based research agent architecture overview
- 34:46Deep research agent design and tools
- 40:01Code walkthrough: MCP server setup and tool registration
- 43:35Deep research tool implementation and prompting
- 45:06Analyze YouTube video tool implementation
- 47:39Compile research tool implementation
- 53:35Live demo: Running the research agent
- 59:56Agent skills and workflow prompting
- 1:10:35Introduction to technical writing workflow
- 1:15:46Writing guidelines and control techniques
- 1:19:06Few-shot prompting for writing
- 1:21:28Evaluator-optimizer pattern (Writer/Reviewer)
- 1:28:16Running the writing post skill
- 1:34:46Observability and tracing with Opik
- 1:41:44LLM Judge implementation and evaluation
- 1:47:56Dataset management and F1 score computation
related talks
32mFrom Deep Research to Evaluation and Optimization of Advanced AI Age...
01h 54mBuilding distributed multi-agent systems
35mBuilding LLM-based Agents
52mCut the Hype and Build Your Own AI Agents
14mUnconference: Self-optimizing deep research agents
35mThe Podcast Pipeline Nobody Sees: Behind the Scenes with AI Agents and AI gen
01h 23mA2A & MCP Workshop: Automating Business Processes with LLMs
49mLet's build an AI agent