A proven process, refined over 35+ machine learning projects 

Our machine learning team has brought dozens of successful projects to market. We’ll guide you through the process to help you minimize your risk and dodge the mistakes that push ML projects off course and over budget.

how-we-work-illustration

Your 4 steps to machine learning success

1

Duration: 1 day

Strategy Day

Pick the right problem

Your project begins with an intensive on-site Strategy Day.

It’s an exploratory, full-day session where we get to know your business in depth. We help you define the right problem to start with and choose the best approach based on your business, domain, goals, and data.

Preliminary validation of the solution helps to avoid costly missteps— like asking the wrong question, lacking sufficient data or choosing an overly complex problem.

2

Duration: 4 weeks

Proof of concept

Validate your solution

We leverage our engineering skill, experience, and pre-built compute infrastructure to arrive at PoC for your ML solution in 4 weeks or less.

Getting to PoC quickly and affordably is essential to get buy-in from your team early in the process. And to minimize the risk of a production solution that falls short of expectations.

At this stage, we validate the algorithm to be sure that it solves your problem. And then make adjustments or redirect our efforts based on what we’ve learned.

3

Duration: 2-6 months

Develop, integrate, test

Your ML solution goes live

Our team works in 2-week sprints to develop your solution, with bi-weekly calls to discuss our progress. Real-time communication via Slack or MS Teams keeps issues moving forward and helps bypass communication bottlenecks.

Then we take the solution live to evaluate its performance in your production workflow.

Finally, we automate everything, fine-tune the details, and hand over your solution— congratulations, your data is now officially working for you! 

4

Duration: Ongoing

Learn. Scale. Grow.

Build on your success

A successful ML application isn’t an endpoint— it’s a launching pad. So what’s next? 

We can help you quickly scale your solution to other locations, projects or products. Or build completely new strategies for helping to slay inefficiencies, improve processes or boost revenue.

Or your engineering team can take it from here and start building on what they’ve learned. We’re here to consult when your team gets stuck.

client-logo

We are very pleased with the support provided by Data Revenue - in particular their knowledge, and the application of machine learning, project management, and client interaction.

,

What to expect from us

communicating-icon

Clear communication

We don’t have a ‘sales team.’ Or ‘junior’ engineers. Or account managers. You work directly with our A-team on every project. 

attention-icon

All-access, all the time

We work with a few clients at a time, so you have our full attention. The CEO & CTO manage your project personally and know every detail inside and out.

price-icon

No counting hours

You never get a bill from us for hours worked. We agree to a project price and that’s what you pay. No cost overruns. No overtime.

learn-icon

Learn from our team

We work alongside your team and they get full access to our code throughout the project. So your team can learn from us to level up their ML skill set.

What you get

Proactive project management and a custom-built solution that works:

planning-icon
Project planning 

Expert interviews, data analysis, and academic research

optimization-icon
Optimized algorithm

Feature engineering, algorithm selection, and tuning

documentation-icon
Clean, tested & documented code

Keep it, use it and repurpose it as needed

parallelization-icon
Infrastructure and parallelization

From thousands to billions of decisions per day

import-icon
Integration into your tool-chain

Auto-importing data, exporting results, or API accessibility. 

transparent-icon
Complete transparency

Full access to plans, issues & our git repo

management-icon
Project management

We keep the project on schedule and you in the loop

updates-icon
Frequent updates

Weekly engineering calls & bi-weekly milestone presentations

communicating-icon
Real-time communication

Slack or Microsoft Teams to keep things moving

client-logo

The team at Data Revenue is a pleasure to work with – open, curious, professional, and non-bureaucratic. They take an active role in every step of the process, and they think ahead. They went the extra mile for us, and we really trust them.

Joern Hagenguth

,

Head of Relocation

Peek inside our toolkit 

We use tools that allow for rapid iteration, reproducibility, easy scaling, and solid architecture. These are the standard building blocks of a Data Revenue machine learning solution:

toolkit-logo
Python
-
Our main development language.
toolkit-logo
Pandas
-
Great for exploration and feature engineering.
toolkit-logo
Dask
-
Leaves Spark in the dust.
toolkit-logo
Luigi
-
For managing all tasks in an execution graph – like Airflow.
toolkit-logo
AWS
-
One stop for compute infrastructure.
toolkit-logo
Google Cloud
-
Great platform as well.
toolkit-logo
Microsoft Azure
-
Azure is fun too.
toolkit-logo
Docker
-
Portable, flexible and simple deployments.
toolkit-logo
Kubernetes
-
Faster, more efficient and agile infrastructure.
toolkit-logo
XGBoost
-
Gradient boosting powerhouse.
toolkit-logo
scikit-learn
-
Grab-box of algorithms and more.
toolkit-logo
Prophet
-
Forecasting library for GAMs – from Facebook.
toolkit-logo
Zipline
-
Pythonic algorithmic trading library.
toolkit-logo
LightFM
-
Hybrid recommender tool.
toolkit-logo
Flask
-
Web apps & APIs.