Bring us a real problem. We'll bring the research.
Analyticity is an applied AI research company in its early chapters. We're building Skrrol and Unitly alongside a handful of design partners — teams willing to share a real workflow, work with us weekly, and let the product be shaped by the problem instead of the other way around.
Three domains where better reasoning changes an outcome.
We're focused. The program exists to put Skrrol and Unitly in front of problems where unreliable AI is worse than no AI, and where a real system — evaluation, latency, governance, the whole stack — wins.
Content platforms & publishing
Short-form video understanding, moderation, and retrieval at editorial scale.
A publisher, creator platform, or media-tech team with a large video corpus and a real trust-and-safety workflow.
Scientific & engineering software
Unit-aware numerical reasoning for simulation, measurement, and multi-module data pipelines.
A scientific-computing, climate, energy, or engineering team where silent unit-conversion bugs have a real cost.
Healthcare & regulated numerics
Dosing, lab-result interpretation, and any workflow where a decimal-place error is a safety incident.
A clinical informatics, pharma, or medical-devices team with a concrete numerical-reasoning gap.
Specific commitments, both directions.
Design partnerships fail when expectations are vague. Here is exactly what we commit to, and exactly what we need back.
- Direct access to the research team
Weekly working sessions with the engineers building the product — not a customer success layer on top.
- Co-authored evaluation harness
We build the eval suite together against your real data distribution, not a generic benchmark. The suite is yours to keep.
- Roadmap input
Design partners get first look at new capabilities and meaningful influence on what ships next.
- Preferred pricing through the first production year
Design partners commit when the work is hardest and the product is least finished. The pricing reflects that.
- A real, scoped problem
Not a science project. A workflow where better AI reasoning directly changes an outcome your team already cares about.
- Weekly working sessions
One hour a week with someone on your team who knows the problem deeply. Partnership requires partnership.
- Evaluation data
A representative slice of your workload we can evaluate against — anonymized where needed, contracted always.
Questions we hear from prospective partners.
How selective is the program?
We're taking a small number of design partners per quarter across Skrrol and Unitly. We'd rather work deeply with a few teams than thinly with many.
Do we need an in-house ML team?
No. You need a team that understands the problem domain. We bring the ML. What we need from you is substantive access to the workflow we're trying to improve.
What happens after the design-partner period?
Partners move onto commercial pricing on favorable terms. The eval harness and model cards we built together stay with you. Public references are opt-in, never required.
Can we keep our work confidential?
Yes. Every partnership is covered by mutual NDA. Case studies, logos, and named references are opt-in and signed off by you before anything is published. We'd rather have no case study than an unhappy partner.
We'd rather earn a reference than publish a logo.
If a problem in your workflow has the shape we're looking for, we want to hear about it — even if you're not sure it's a fit. The application is short; we reply to every serious inquiry.