01
Assess
Four weeks with your analysts and engineers. We map every source, every report nobody trusts, and what each gap costs in hours.
- Source and quality inventory
- Ranked opportunity list
SOC 2 Type II · EU AI Act ready · ISO 27001
Data and AI consulting for large organisations
We build the platform, the models and the governance underneath them, then hand the whole thing to your team.
- Founded
- 2016
- Models in production
- 190+
- Pilots that shipped
- 78%
- First value in
- 6 wks
Trusted by data teams at
What we do
Eight practices, one accountable team
No offshore handover. The engineer who designs your model layer is the one who sits in your sprint review.
Data strategy
Audit of what you have, a target operating model, and a roadmap with a business case your CFO will actually read.
Read moreData engineering
Warehouse migration, ELT pipelines, dbt modelling and streaming, built to be read and changed by your own engineers.
Read moreAnalytics & BI
A semantic layer where every metric has one definition, one owner and one place it is calculated.
Read moreMachine learning
Demand forecasting, churn and credit scoring, anomaly detection and predictive maintenance, scoped one decision at a time.
Read moreGenerative AI
Retrieval over your own documents, support and sales copilots, document extraction, with evaluation and guardrails from day one.
Read moreMLOps & platform
CI/CD for models, feature store, drift monitoring and cost control on the platform you already run.
Read moreGovernance & AI Act
Catalogue, lineage, data quality and ownership, plus risk tiering and model cards that satisfy an auditor.
Read moreTraining & enablement
Workshops, a centre of excellence, and coaching for the analysts who take the work over.
Read moreFrom first audit to a team that runs it alone
Where you are now
Take the assessmentFive levels of data maturity
Most organisations that call us sit at level two and believe they are at level four. Find your row honestly, then read the one below it.
L1
Spreadsheets
Numbers are assembled by hand each month. The reporting cycle depends on two people and their files.
L2
Reports, no agreement
Dashboards exist, but each team built its own. The same metric is calculated three ways and nobody owns the definition.
L3
One source of truth
A warehouse with modelled, tested data and a shared metric layer. Analysts answer questions without asking engineering first.
L4
Decisions automated
Models run in production against live data, monitored for drift, with a clear owner and a rollback path for each one.
L5
Governed and auditable
Lineage, access and model documentation are current by default, so an audit question is answered the same day it is asked.
Selected work
Four organisations, four very different questions
Figures below are the client's own, measured ninety days after handover.
Ridgeline Bank · Columbus, OH
A month-end close that took eleven days
Finance rebuilt the same figures by hand every cycle. We modelled the ledger in the warehouse, defined forty metrics once, and automated the reconciliation.
11 → 3
Days to close
40
Metrics defined once
Carter Foods · Lima, OH
Forecasting demand for 2,400 SKUs
Planners worked from last year plus intuition. We built a forecasting service per SKU and region, retrained weekly, with a manual override planners actually use.
-19%
Forecast error
$2.1M
Working capital freed
Allegheny Health · Pittsburgh, PA
Twenty years of clinical notes, finally searchable
A retrieval system over de-identified records, with citations back to the source document and an evaluation set reviewed by clinicians each quarter.
6 min
Median search time
100%
Answers with citations
Hollis Logistics · Kenosha, WI
A warehouse bill that doubled in a year
Nobody could see which queries cost what. We added cost attribution per team, rewrote the twelve worst models, and set a monthly budget alert.
-46%
Monthly platform cost
5 mo
Payback
About GridlineRead our story
We would rather be unnecessary in a year
Gridline was started in 2016 by three analytics leads who were tired of consultancies that left behind a platform nobody internally could change. Forty-two people now, employee-owned, still working that way.
Everything we build lives in your cloud account and your repository from the first commit. No proprietary layer, no licence that follows you, no dependency designed to renew itself.
Employee-owned
No outside investors
SOC 2 Type II
ISO 27001 certified
EU & US delivery
Data residency on request
Your code
Open stack, no lock-in
We know your regulators, not just your schema
Platforms
We work in your stack, not ours
Certified on the platforms below. If you have already chosen, we fit in. If you have not, we will tell you which one costs least for your workload.
Snowflake
Warehouse design, cost attribution, secure data sharing.
Databricks
Lakehouse, Unity Catalog, MLflow and streaming jobs.
Azure & AWS
Landing zones, networking, identity and cost guardrails.
dbt & orchestration
Modelling standards, tests, CI and scheduled runs.
They told us our second use case was not worth building. That one sentence bought them every project since.
1 / 3
91%
Clients who return
42
Consultants on staff
190+
Models in production
2016
Year we started
The people who answer your call
Ray Novak
Co-founder & Chair
Erin Castellanos
Chief Executive
Devon Pruitt
Head of Data Engineering
Tasha Bramley
Head of Machine Learning
Owen Mbeki
Head of Platform & MLOps
Lauren Fisk
Head of Governance & AI Risk
Questions
What buyers ask us first
Still unsure? Call (216) 123 45678 and ask for whoever is on the desk.
Ask a questionYou do, from the first commit. Everything is built in your cloud account and your repository, on open tooling, with no licence that follows you after we leave.
No. Client data is never used to train shared models, never leaves the region you specify, and is covered by a DPA plus named subprocessors published in our trust centre.
Every system we build is risk-tiered at design time and ships with a model card covering training data, evaluation, limitations and human oversight. High-risk use cases get a documented review before any code is written.
We agree the success threshold before starting. If we miss it, we write up why, hand over what we learned, and recommend against scaling. That has happened four times and we would do it again.
Almost always. Your engineers review our pull requests and ours review theirs. If your team is not involved, the handover fails, so we scope it in from the start.
Tell us which number you cannot trust
Send one paragraph about the report that always starts an argument. A data lead, not a salesperson, replies within one business day.