Why now
Representation is practical.
Embeddings and similarity search make exploratory science tractable in software — discovery layers can sit beside conventional predictive stacks.
Investor thesis
Modex is building a cross-domain discovery engine for hidden structure in real-world systems.
The thesis is that valuable signals remain underinterpreted because tooling is fragmented or domain-locked. Modex is an experimental ML platform for ingesting, embedding, comparing, clustering, and hypothesizing across unfamiliar systems — starting from a credible experiment console and growing into richer modalities.
Wedge
Experimental signal discovery.
Initial focus: typed experiments across animal signals, fluids, materials, sequences, and unknown spaces — before locking into a single vertical.
Moat path
Workflow + data flywheel.
Defensibility grows through domain-specific flows, proprietary datasets where permitted, cross-domain tooling quality, and UX depth as usage compounds.
Business shape
Research tooling & APIs.
Expansion vectors include research workspaces, enterprise experiment pipelines, vertical modules, and API access to discovery primitives — timed with validation rather than hype.