Botkeeper was my introduction into the world of startups, when I was brought on as the Machine Learning Engineering Team Lead back in 2020. Since then, I've been able to support them with machine learning and data engineering contributions, both as a full time employee and as a consultant. Being able to be a part of their early phase evolution and continue to have a front row seat to watch them grow, has been really special.

This company is incredibly focused on data, with a product that uses machine learning to assist in the bookkeeping process. In order to support the range of general ledger software that is popular with their clients, they require data normalization capabilities, robust data pipelines, and highly customized machine learning models.
‍
When I first started with Botkeeper, the model was in its first iteration and only supported a single general ledger. I was the only full time employee on this new Machine Learning team, and I had set some ambitious goals for my first six months: I wanted to improve the models performance, make architecture adjustments to support increased model complexity, externalize model storage, automate training, and support an additional general ledger by replacing the existing data pipelines with new ones that involved data normalization. Loving my newfound freedom of green fields, having the support of an amazing engineering team, and having experts on staff who could help explain the domain at low levels, I was able to accomplish all of these goals.

In the years that followed, we we hired other ML and data engineers. We were able to further improve the model through transfer learning, and support additional operational processes with machine learning. As the model's complexity increased, so did our infrastructure and design requirements. We moved towards asynchronous push patterns with decoupled workers and auto-scaling in order to support both on-demand and large nightly batch consumers. With these architecture adjustments, we were able to maintain a service of high reliability and efficient use of hardware.

My role evolved to include managerial, mentoring, and project management responsibilities as I began to plan the roadmap for ML and data engineering for a full fledged team. I continued to contribute code in both areas, but was doing more architecting and delegating. With a maturing team and a roadmap set for some time out, I transitioned into full time consulting. I was incredibly lucky that Botkeeper wanted to maintain a relationship with me, and that I was able to continue to support them and work with the team.

In the consulting role, I have been able to continue to offer architecture and mentoring services, as well as contribute to ongoing projects. In the year and a half that I have been supporting them part time, we've been able to implement a lot of great data engineering features, such as: a new data normalization layer that enables onboarding of additional general ledgers, improved reliability of ETL solutions at scale, and reduced redundancy in code through extensible patterns for generic data processes and normalizing data further upstream. I've also been able to continue to support the ML engineers with their modeling efforts, and have loved watching them come into their own and thrive as leaders.

I love working with this company and its team members and cannot wait to see what they build next!