Scaler AI Labs is now Evaratus AI
Evaratus Reach out athello@scalerailabs.com
RL Environments · Evals · Enterprise Data · Applied Research

Real work, made trainable

Reach out at hello@scalerailabs.com
01 - Mission

Free humans from mundane work

02 - Problem

Frontier models still fail at real-world tasks. We fix it

03 - Apparatus

AI needs continuous research, evals, gyms and data to become effective at work - we are the apparatus that makes it happen.

4 / 5Top frontier labs work with us
$30M+Of work delivered every monthAs of August 2026
250+Enterprises as exclusive partners
MillionsOf experts from our community & partnersThrough our own community and the userbase of our enterprise partners — engaged exclusively with partners
Our differentiator

Foundational Pillars

that are hard to copy
12 34

Enterprise data

The largest repository of proprietary enterprise data.

Fundamental to any work that will happen in AI learning for enterprises.

  • Workflow and domain-specific data from 350+ sector leaders
  • Sourced at scale - pre-IPO and IPO’d companies - workflow data complexity increases with company size and age
  • One of the largest intelligence corpora across regulated spaces - healthcare, finance, law and more
  • PII/BII scrubbed but workflows retained - SOC 2 compliant
HealthcareFinanceEcommerceEducationManufacturingSupply ChainInsure-TechDevToolsCoWorkingHealthcareFinanceEcommerceEducationManufacturingSupply ChainInsure-TechDevToolsCoWorking

Consumer data

One of the richest and largest proprietary consumer datasets.

Essential for building out personal superintelligence.

  • Sourced directly from power users across different usecases
  • Health tracking, travel planning, tax filing, finance management and many more
  • Covering 25+ countries and 100+ usecases
  • PII/BII scrubbed while retaining workflows - SOC 2 / GDPR compliant
Health TrackingTravel PlanningTax FilingFinance ManagementFitnessParentingGamingLearningShoppingHealth TrackingTravel PlanningTax FilingFinance ManagementFitnessParentingGamingLearningShopping

RL gyms

The largest collection of high-fidelity computer-use and tool-use RL gyms.

Where models practice real work before they do it in the world.

  • 1000+ enterprise and consumer RL gyms
  • High-fidelity workflow mapping from domain experts - on every gym
  • 1000+ multi-app gyms simulating complete jobs
  • Built to the shape of the job - not to what is easy to verify
Team ChatImage EditorCloud StorageCADContract ManagementTask ManagementSales CRMIT TicketingSupport DeskEHRMulti-app Finance ManagerMulti-app Product DesignerTeam ChatImage EditorCloud StorageCADContract ManagementTask ManagementSales CRMIT TicketingSupport DeskEHRMulti-app Finance ManagerMulti-app Product Designer

Expert network

10M+Experts across
Software engineers
Data scientists & MLEs
Medical practitioners
Tax consultants
Investment bankers
Lawyers

Together, these form the apparatus for model uplift

The four pillars - enterprise data, consumer data, RL gyms and an expert network - converge into Evaratus' applied research loop, where failure analysis, task design, verification and reward modeling run as repeated evaluation passes. Performance is measured on held-out tasks before and after training; the difference is the measured uplift. Changes ship only after that uplift is proven. Deployed models then surface new failures, which become new tasks and new data for the next cycle.

00 · Proprietary inputs
  • Enterprise data
  • Consumer data
  • RL environments
  • Expert network
01 · Applied research
01 Failure analysis02 Task design03 Verification04 Reward modeling
02 · Held-out evaluation
Baseline→Post-training
Δ
Measured uplift
Ships only after proven uplift
Deployed models surface new failures
→ new tasks → new data
Evaratus · Held-out eval · The apparatus
01 - Origin

Evaratus AI Lab is built by the same team that built Scaler to a $710M valuation, with people who’ve shipped at Meta, Google, Microsoft, Goldman Sachs and Sprinklr.

02 - Apparatus

We build the training and evaluation apparatus for frontier AI: RL environments, evals, enterprise data and applied research. 4 of the top 5 frontier labs work with us.

03 - Proof

We work exclusively with our partners, and nothing ships on promise: data, evals and RL environments ship only after we prove model uplift on held-out tasks.

Team

Built by people who've

shipped at scale
Anshuman Singh
Founder & CEO

Anshuman Singh

Built Scaler to a $710M valuation. Two-time ACM-ICPC World Finalist, former Facebook engineering lead for Messenger/Chat and helped establish its London office.

MetaScaler
Naman Bhalla
Founder/Research

Naman Bhalla

Ex-Google. Led AI Infra at Google. Scaled AI products to 5M+ users at Scaler. Early member of an Indian unicorn.

GoogleScaler
Mohit Srivastav
Product/GTM

Mohit Srivastav

Built Zeta’s (unicorn) finance stack; ex-Goldman Sachs, ex-Amazon.

Goldman SachsAmazonMicrosoft
Aishvarya Singh
Research

Aishvarya Singh

ICPC World Finalist. Stanford AI. A decade of experience across Google, Meta and Apple. Part of the research and infra team at Apple.

MetaGoogleApple
Venkateswara Rao Ch
Research

Venkateswara Rao Ch

2x ICPC World Finalist. IIT-B CS. Research at Microsoft, Google. Founding engineer at 0xppl and Composio.

GoogleMicrosoft
Vikram Aditya
Product/GTM

Vikram Aditya

Ex-Product and GTM leader at Sprinklr. Ex-founder, raised $5M+, built 2 AI companies.

Sprinklr
Previously shipped at

Build the apparatus

with us
Reach out at hello@scalerailabs.com
Applied Research

Areas of

Research
Coding

proven deltas

Data, evals and RL environments ship only after we prove model uplift on held-out tasks.

Reach out at hello@scalerailabs.com