We built Tempo, to accelerate the rhythm of science and business development. An agent-based AI research assistant that converts any research prompt into a fully formatted, 20-page, academic-style PDF report with all sources cited. The goal was to democratize access to deep research by making high-quality, PhD-level insights available on demand.
Tools: OpenAI GPT-4o – Main content generation and initial draft Claude 3 – Fact validation and argument coherence LangChain – Agent orchestration and memory logic Parallel - Taking many different sources and coalescing them into one output
We did:
- Four hours of phone calls with real customers who were researchers from JH, MIT, BU
- Agent cross-verification reduced hallucination and surfaced accurate insights
- Automated PDF composition with references allowed for quick professional output
- Verifiable citations with complete sources and references page at the end of each report
- Securing an REAL LOI from a Diego Maye who was willing to pay $500 a month
Some things that were challenging:
- Ensuring citation accuracy across multiple independent data fetches
- Getting real researcher to want to use the product
- Enabling the creation of graphics on the end output like slides, and linkedin post, etc
Log in or sign up for Devpost to join the conversation.