AutoResearcher
Seven-agent research-report workflow
- Status
- Course capstone
- Engagement
- Course capstone
- Role
- Full-stack & AI engineer
The problem
Building a single citation-rich research report still takes many hours of manual searching, writing and formatting.
What was built
A seven-agent research workflow built on Google ADK and Vertex AI that transforms a prompt into a print-ready LaTeX PDF with sources, charts and citations.
Architecture and engineering decisions
A hierarchical seven-agent system on Vertex AI ADK using Gemini 2.5 Pro and Flash, with Tavily web sources, claim fact-checking, Matplotlib charts and Pandoc/LaTeX export.
Audience
Professionals and students who need structured research reports.
Technical highlights
- Seven-agent hierarchical system (Vertex AI ADK)
- Gemini 2.5 Pro and Flash models
- Tavily web sources and claim fact-checking
- Matplotlib charts and Pandoc LaTeX PDF export
My responsibilities
- System design
- Agent orchestration
- Notebook implementation
Why it matters for a client
Shows agent orchestration, source grounding and document generation in a constrained research workflow.
Constraints and trade-offs
Educational and capstone submission purposes; sources must be cited accurately and zero hallucinations are not claimed.
Privacy and accessibility
Runs on Kaggle or Colab with user-supplied API keys; no personal data is stored by DevRay.
Verified evidence
Public repository at github.com/rayklanderman/CapstoneProject-Autoresearcher. Described on the owner portfolio as a seven-agent Vertex AI ADK capstone.
Limitations
Capstone scope; not a production service; accuracy depends on the underlying models and search tools.