Too Much

Having problems scaling an existing data-related process? Spreadsheets getting too complicated? Or maybe you have so much data you don't know where to start. I automate time-consuming repetitive tasks and help you decide which opportunities to pursue now, and in the long-term. I can also help you build tools to search through your data, whether you are an analyst bogged down with slow SQL queries or an HR specialist unable to locate a form on your intranet.

Too Messy

Many businesses struggle with poor data quality, from sole proprietors to multi-national corporations. It's hard to make informed decisions when you can't get what you need, when you need it. I help clients assess data quality and how it's impacting their business, improve it with better data collection practices, and make the most of what they already have. Perfect data isn't practical; often times great value can be had from data that's just good enough.

Optimize

Some obvious questions can be very difficult to answer: What price should I charge? Which combination of products in a bundled offer will yield the most profit? Which part of my inventory can I liquidate to free up the most cash with the least risk? These are all questions I can help you answer.

Cutting Edge

Thanks to widely available open source libraries and cheap cloud computing you don't need to hire a team of deep learning experts or buy an expensive proprietary appliance to benefit from some of the unstructured data you probably already have: customer feedback, blog posts, documents, images, audio, or social media posts. Software that can automatically summarize, classify and act on unstructured data is not just for Fortune 500 companies. I also have experience using domain-specific data to produce custom models that perform better at specific tasks than the widely available machine learning APIs offered by major cloud providers.

From Reactive to Proactive

Reporting tells you what happened, real-time reporting tells you what is happening, and forecasting tells you what could happen. By training a predictive model on historical data and conducting time-series analyses, I can help you make informed decisions before your normal reporting would usually be available.

Customer Insight

Some decisions on how to interact with your customers have to be made, regardless of available data quality. Validating the assumptions you make about your customers can be critical. I can help you learn more about your customers and validate the assumptions you have already made. I have experience analyzing unstructured text like customer reviews, chat transcripts, and comments; mining log files for pain points and buying patterns; segmenting them into personas; scoring leads; modeling churn and optimizing retention.

Siloes

Each department has its own database, for its own purposes, often undocumented outside of that department. How do you know if the 'John Smith' in your marketing database is the same customer as the 'John Smith' in your sales database? I have experience partnering with in-house teams to assist in combining disparate schemas, building data dictionaries, ETLs, and matching and merging millions of contact records.

Guidance

There has been an explosion of data-related vendors and services in the past few years, and likewise, there is no shortage of salespeople eagerly telling you how you're a perfect fit for their one-click solution. I provide clients with technical advice that allows them to see through the hype and find the right tools and services for their business. I also have worked with several smaller customers looking for advice on their fist data science hires and efforts.

Dashboards

Custom, interactive, in the browser. Click the GIF below to see an example.

Run ETLs

Customized, easy to use interfaces for running complex automated processes.

Notebooks

Jupyter notebooks allow me to show a client exactly how I performed each step of an analysis and tailor the explanation to the client's technical ability. It is an important tool I use to make data accessible to a wide variety of teams within a client's organization. Decision makers can use them to see specific results and visualizations, while hiding complex details. The detail is easily revealed allowing the client's in-house team to quickly verify the results; even one-off analyses need to be reproducible and will come version controlled with documentation.

Code

For projects that involve more software engineering effort than one-off analyses such as real-time recommenders, classifiers, and ETL’s, I like to adhere to additional best practices to ensure the deliverables are production-ready and maintainable even if I’m not the one doing the deploy. Tests, doc-strings, consistent formatting, logging, and refactoring code into pip installable python packages where appropriate are the norm. Code Sample