Hi all,
I built Genesis — a multi-agent scientific discovery engine written in Go.
You ask it a scientific question. Three agents run in parallel Go goroutines:
- Climate agent → queries Earth-2 / FourCastNet for climate projections
- Molecular agent → queries BioNeMo for protein-level biological analysis
- Reasoning agent → NIM Nemotron-70B synthesizes both into a cross-scale hypothesis
Example:
genesis ask “How will rising sea temperatures affect coral bleaching by 2035?”
What’s different here is the architecture — Go goroutines map naturally to
multi-agent parallelism. Agents push results into a typed channel, no locks,
no shared memory. Single binary, no runtime dependencies.
Three commands to run it:
git clone GitHub - HarshalSant/genesis: Agentic Scientific Discovery Engine — connects NVIDIA NIM, Earth-2 and BioNeMo into a parallel multi-agent pipeline in Go · GitHub
go build -o genesis ./cmd/genesis
NVIDIA_API_KEY=your_key ./genesis ask “your question”
Free API key: https://build.nvidia.com
Would love feedback from anyone working on NIM or Earth-2 integrations.