Nettle • Autonomous Data Processing Platform

NEWS

Sequoia joins forces with Nettle

→

NETTLE

Why

The Engine

Proof

Team

Investors

Try Nettle

Join the Waitlist

Autonomous Data Platform · Live Today

Months of data engineering. An hour of Nettle.

Point Nettle at your raw data. It stands up a state-of-the-art, AI-managed data stack — pipelines, models and decision infrastructure — with no schema docs, no configuration, and no forward-deployed engineers.

Fully autonomous

Self-healing

Zero maintenance

Join the Waitlist

Talk to Us

The engine underneath:

15 subsystems · 120 pipelines

7 models · 45 invariants —

stitched into one control plane,

maintained automatically.

nettle · autonomous data engine

v3.2 — live

Your sources, as they are

s3://crm_export_final_v2.csv

no schema

✓

postgres.public.orders

4 dup keys

✓

kafka:orders.v3

late events

✓

stripe/charges.json

nested

✓

sheets/eu_freight.xlsx

hand-edited

✓

Ask anything

nettle ›

Which accounts will churn next quarter?

› 2 cohorts flagged — renewal risk & usage decay

› action triggers armed for owners

predictive

enterprise_churn

fresh 2s ago

Structured and queryable in about an hour. No pipeline written.

<$5 LLM cost per dataset · <$35 compute per day

98%

classification accuracy over unseen datasets

90%

less engineering time on setup & maintenance

90%

lower token cost per data use case

10×

data-team productivity, immediately

// measured across 50 tables in 13 datasets · Q2 2026

The Problem

Every era promised to end this. Every era added a layer.

The warehouse didn't remove the ETL. The stack didn't remove the warehouse. The copilot didn't remove the pipeline. Each one arrived with its own headcount — and enterprise AI is now stuck behind all of them.

One engine. Not one more layer.

1995 Data warehouse

+ a team to own it

2005 ETL & orchestration

+ a team to own it

2015 The modern data stack

+ a team to own it

2023 Copilots on top

+ a team to own it

systems to maintain

50% of software payroll

works on the data layer or the systems around it. The toil lives where the data lives.

70% of AI token spend

burns at the same layer — agents re-reading raw, unmodelled data on every single run.

$5.5B+

of announced capital has gone into hiring humans to sit next to enterprise data. The bottleneck of the AI revolution is the supply of forward-deployed engineers — and Nettle removes the need for them.

Legacy data isn't AI-ready

Decades of schemas, exports and hand-edited sheets that no model can safely reason over.

Agents aren't in the core loop

AI still isn't driving core business processes — it lacks context data that is current, correct and maintained.

Point solutions miss the picture

Tool-by-tool fixes patch one pipeline while the surrounding hairball keeps growing.

And a layer is never just software. It's the bill.

Headcount, on-call, and a rewrite every time the business changes.

That is where the engineering budget actually goes.

What your team owns today

Ingest and connector code

Transformation logic, per decision

Schema migrations and backfills

Reconciliation when sources disagree

The rewrite, every time the model changes

…and one more line every time the business asks a new question.

What your team owns with Nettle

✓

The questions worth asking

// that's the list

Your engineers stop maintaining the answer machine — and go build the product.

Millions in platform spend drops to the cost of electricity:

<$35 compute per day · <$5 LLM cost per dataset.

The Engine

One plane underneath.

Every decision on top.

Scroll — the engine assembles itself.

15 subsystems

120 pipelines

7 models

45 invariants — maintained automatically

Raw Data

Autonomous Data Plane

Semantic Layer

Apps

Explanatory

Causal

Models

Visualisation

OUT

Decisions & actions

01 · Raw Data

Point it at what you already have.

No schema. No documentation. No pipeline written first. Legacy sources aren't a blocker — they're the input.

Postgres

Kafka

SaaS APIs

Files

400+ connectors

02 · The Foundation

It models itself.

Ingests, cleans, indexes and self-heals continuously. Nothing above it ever touches raw data.

01Entities & keys resolved

✓

02Types & formats normalised

✓

03Duplicates & late events reconciled

✓

04Watching for drift

●

// nobody wrote any of this

03 · Semantic Layer

Your business, in your own terms.

Entities, metrics and rules inferred from how you actually operate — then kept in sync as you change.

customer

order

margin

churn

cohort

+ whatever you add next

04 · Apps

Two apps live. The rest, one by one.

Every app sits on the same plane and the same semantic layer. Unblock one, and the next costs almost nothing to add.

Explanatory

BUILDING NOW

Talk to raw data. Insight and reporting, without a dashboard being built for you.

Causal

BUILDING NOW

Not what happened — why. Isolate the driver, size it, decide what to change.

Models

Predictive models fitted on the semantic layer — churn, spend, vendors — refit whenever it moves.

Visualisation

Views that assemble themselves from the questions being asked.

Or your own.

Your engineers build agentic and data apps against entities and metrics — never against raw data, processing or cleaning.

REST

SQL

MCP

streams

OUT · Decisions & Actions

Decisions, not dashboards.

Forecasts, diagnoses, alerts and agent actions — emitted continuously, in the terms the business already uses. Serving 16 personas — exec, finance, legal, sales, ops and the rest — without a report being written for them.

Then the business changes — and none of this breaks.

Something moves

Drift detected

Engine re-models

Decisions stay true

Where We Are

Live today. Building in the open.

Phase 01 · Live Today

Ingest to insights, without being told anything

Nettle connects, understands the structure and returns insights — no schema documentation, no configuration. Your team can always override its choices.

Natural-language querying & SQL

Semantic classification

BI in one environment

Phase 02 · In Development

Autonomous data migrations

One direction of flow: your sources go in, a complete platform comes out. Generated in about an hour, instead of built over months. Nothing authored by hand in between.

Silver & gold tables (Iceberg)

Realtime OLAP · search · streams

BI / MOLAP · semantic models

Phase 03 · Roadmap

It runs the platform itself

Three things happen continuously — monitored against the model, not a runbook.

Run

— pipelines & invariants execute

Heal

— drift reconciled automatically

Build

— new sources & questions absorbed

The Technical Advantage

A data model borrowed from physics.

Most platforms leave the data model up to each business — that is how you get today's mess. Nettle's models closely track differential geometry, quantum field theory and group theory, organised to solve for business decisions.

A stable model is why the engine works on industries and datasets it has never seen — and why predictive models can be built repeatably over them.

Differential Geometry

Quantum Field Theory

Group Theory

Loop Quantum Gravity

String Theory

10+ PhDs

Modelling group drawn from MIT, Fermilab, Brown and Northwestern — R&D complete, advising as needed.

Proof

55 to 98, in two quarters.

Table and column classification accuracy over unseen datasets — the capability every autonomous decision above it depends on.

// 50 tables across 13 datasets

// no human in the loop on the onboarding path

55%

88%

98%

March 2026

first working prototype

April 2026

first optimisation pass

Q2 2026

across unseen datasets

Why Now

The autonomous data layer is being defined right now.

Demand already exists

Every enterprise AI deployment needs context data that is current, correct and continuously maintained. That is a data-layer job.

The category is forming

In the last three quarters Google, Microsoft, Oracle, Databricks, Snowflake and Fivetran + dbt all announced agentic data roadmaps. Palantir has a head start.

Depth wins

Revenue will flow to the most complete and sophisticated solutions — not the widest patchwork.

Complete and autonomous:

the corner nobody occupies.

Copilots assist. Hyperscalers assemble. Point solutions patch. Nettle is the only unified platform that deploys and operates with no human in the loop — which is what makes adoption fast where FDE-led models stay slow.

UNIFIED PLATFORM →

POINT SOLUTIONS

FULLY AUTONOMOUS →

In-house + LLM patchwork

Point solutions

Assistive / copilots

Hyperscalers

Nettle

The Team

Operators and physicists.

Divye Kapoor

Founder & CEO

Mohit Kumar

Founding Product Manager

Alexey Sanko

Founding Engineer, Data

Nitin Shukla

Founding Engineer, AI

Francisco Franco

Founding GTM & Sales

Michelle Teo

Founding Growth & Marketing

Andrea Jane Caluma

Founding Business Operations

We're hiring

Engineering · GTM

Board of Directors

Operators from Google, Uber and Pinterest.

Mathematical Modelling Group

10+ PhDs across mathematical physics, quantum information theory and distributed computing.

Design Partners

Bring us your hardest data problem.

We take on a few partners at a time and solve the one problem costing you most.

You bring

The problem your team keeps rebuilding around.

We deliver

A working engine on your raw data, end to end.

We ask

Feedback, data access, and a testimonial — once it lands.

Join the Waitlist

Talk to Us

Invest in Nettle

NETTLE

The autonomous data engine.

Raw data in, decisions out.

Connect with the team

Visit our LinkedIn

Product

The Engine

Explanatory

Causal

Data Plane

Company

About us

Team

Careers

Investors

Legal

Terms & conditions

autonomous data platform · for enterprise