A multi-agent Gemini system that automates deep market research—collecting evidence, cross-checking sources, and producing actionable competitive and industry insights.
Turning “understand this market” into minutes instead of weeks with the help of multiple Gemini Models.
What it does
Five-step pipeline: industry in → taxonomy, segments, behavioral, competitive, Decision Jury → structured report (PDF/HTML); each agent gets only its direct input + summary.
How we built it
Stack (Python, Streamlit, Gemini, Google AI Studio, Pydantic); Gemini Pro vs Flash; five agents + orchestrator + report builder.
Challenges we ran into
Reliable JSON from the model (fences, trailing commas); keeping agents focused; rate limits (429) and retries; Jury prompt size.
Accomplishments that we're proud of
Full pipeline from one string to report; clear agent separation; robust Gemini integration; usable UI and downloads.
What we learned
System instructions + JSON format; Pro/Flash tradeoff; isolating context to reduce hallucination
What's next for Agentic Deep Market Research
Adding a dedicated GTM agent (or agent phase) that consumes the research artifact and produces actionable go-to-market strategy positioning, channels, pricing signals, launch sequence, and success metrics so the pipeline goes from market understanding to execution playbook.
Log in or sign up for Devpost to join the conversation.