NOVA

Paper · Oral

Paper · Oral

Learning
without labels.

A self-supervised method that closes the gap.

+4.2

Points over SOTA

+4.2

Headline metric

Research conference

Prepared for

01 / 20

The problem

Labeled data is the bottleneck of the field — our method matches supervised baselines with 5% of the labels.

machine learning teams feel it every week: the status quo scales cost, not outcomes. There is no clean upgrade path — only rip-and-replace.

02 / 20

Market

01

+4.2

Points over SOTA

02

5%

Labels used

03

3

Benchmarks

Bottom-up, the reachable market clears a nine-figure ceiling before expansion.

03 / 20

[ FULL-BLEED · machine learning hero image ]

Fewer
labels.

04 / 20

How it works

STEP 01

Connect

Map the whole machine learning surface in under an hour. No migration.

→

STEP 02

NOVA core

Golden paths become policy, enforced at runtime.

→

STEP 03

Deliver

Teams self-serve safely with audit and rollback for free.

05 / 20

The product

One surface your whole research conference lives in.

01

Live graph

Every dependency, owner, and cost in real time.

02

Policy-as-code

Guardrails that block bad changes automatically.

03

Attribution

Every dollar tied to a team and a decision.

[ product UI screenshot ]

06 / 20

Traction

Points over SOTA, last 8 quarters

+4.2

+246% YoY

25

50

75

100

113

149

203

256

322

386

480

619

Q3·24Q4·24Q1·25Q2·25Q3·25Q4·25Q1·26Q2·26

07 / 20

Why now

Three forces converging

01

Demand

machine learning budgets are shifting from headcount to platform.

02

Technology

The primitives finally exist to do this at the edge.

03

Distribution

A bottom-up motion the incumbents can’t run.

08 / 20

Adoption

Active accounts

M1M3M6M9M12M15M18

actualguide

09 / 20

Landscape

NOVA

Legacy A

Point tool B

DIY

Breadth → ↑ Depth

10 / 20

Two

Part Two

Ablations.

What matters, what doesn’t, and where it breaks.

11 / 20

Detail

Unit economics

This yearLast yearYoY

Revenue619386+40%

Gross margin78%71%+7 pts

Net retention134%119%+15 pts

Payback11 mo17 mo−6 mo

12 / 20

Where it goes

Use of funds

01

Engineering & product55%

02

Go-to-market26%

03

Operations & reserve19%

13 / 20

Cohorts

Retention by cohort

M0

M3

M6

M9

M12

Jan

100

88

79

72

68

Apr

100

90

82

76

—

Jul

100

91

85

—

Oct

100

93

—

LOWHIGH

14 / 20

Roadmap

The next 18 months

Now → Q2

Land

Own the machine learning beachhead with a self-serve motion.

Q3 → Q4

Expand

Enterprise controls, SSO, and the partner program.

2027

Platform

Open the API and the marketplace.

15 / 20

“

In their words

“A clean result with careful ablations.”

Reviewer 2

Peer review

16 / 20

Revenue mix

Where the money comes from

0100%

55%

26%

19%

Platform 55%

Services 26%

Marketplace 19%

17 / 20

Team

Built by operators

[ photo ]

Reviewer 2

Co-founder & CEO

[ photo ]

A. Rivera

Co-founder & CTO

[ photo ]

M. Osei

VP Product

[ photo ]

L. Chen

VP Revenue

18 / 20

In one number

+4.2

points over the prior state of the art on the standard benchmark.

19 / 20

NOVA