Home

Unit21

San Francisco, USA · Oct 2020 – Feb 2024

This round of my search was focused on exploring the Bay Area, which, despite Covid, was still the center of the tech world. On the one hand, I had a down-leveled SDE1 offer from Amazon. On the other, Unit21 had just raised its Series A.

With Unit21 in a familiar FinTech space and Clarence being one of the most genuine hiring managers I’d come across, I thought I’d learn more joining this startup ride. What a ride it was.

The Work

Software Engineer (Oct 2020 – Mar 2021)

Python Flask PostgreSQL AWS Celery Elasticsearch

Unit21 was less than 30 people when I joined, building a vertically integrated compliance/fraud detection platform. The product & team were broadly divided into three parts - data ingestion, anomaly detection, and alerts investigation.

I jumped straight into the core detection engine. We were onboarding a very large customer at the time, so I worked directly with one of their analysts and built a few custom features for them.

Learning from these examples, I came up with a more extensible design of the rules engine, dubbed “Dynamic Model Builder” (DMB), and made a prototype of it with Caleb over the quieter holiday season. DMB’s more modular design not only supported many more different types of rules for customers, but also came with tangible performance benefits. As a result, it became an immediate hit, both internally and with new customers.

I wrote more about developing the DMB in this slide deck.

Tech Lead Manager (Mar 2021 – Apr 2023)

Python FastAPI PostgreSQL Kafka GitHub Actions Looker

As DMB was gaining popularity, a team started to form around it. Naturally, I stepped up to be the manager, while continuing to be involved technically.

Over the next ~2 years, the team grew from just Caleb and myself to ~15 people at its peak. We took on the entire detection aspect of the product, and worked on many improvements:

  • DMB expansion & roll-out, scaling to >90% of executions
  • Real-time detection
  • Entity risk scores
  • Rule analytics, import & export, bulk actions
  • Rule scheduling & execution optimizations

Besides the day-to-day team management responsibilities, I spent a lot of time on hiring. In the craziness of the Bay Area labor market of 2021, competing for talent as a startup was extremely difficult. There were weeks where I had 15+ phone screens, among other interviews.

Through many iterations of the hiring process and hard work, I managed to convince a few great engineers to join us. On top of it, a few of my prior co-workers & friends also took the plunge with me.

For quite some time, Michael, Peter, and I were the only layer of engineering management at Unit21, all reporting directly to Clarence, the CTO, and we were meeting weekly. I used this forum to build strong org-wide alignment on a number of initiatives, including:

  • The engineering hiring process & interview bank.
  • An early version of the incident management process & post-mortems.
  • The foundations of an org structure based on self-contained product teams, layered with cross-team guilds on top.
  • The Architecture Review Board, where important architectural changes were more formally reviewed by a rotating group of senior engineers.

Staff Software Engineer (Apr 2023 – Feb 2024)

Python FastAPI Snowflake Databricks Stream Processing

As Unit21 acquired more and bigger customers, it became clearer that the data infrastructure on a single Postgres database was getting close to its limits.

We recognized the problem and started experimenting with various solutions as early as 2021. By 2023, with a stronger sense of urgency, and a more complete and validated solution taking shape, the project kicked into a higher gear. With my team hitting its stride on its own, I volunteered to officially tech lead this large-scale data infra migration, coming back to an IC role.

Over the course of the migration, almost all engineering teams — 35+ engineers in total — got involved at some point. I was jumping between the mode of looking ahead — planning & clearing the technical blockers for the next milestone, and the mode of diving deep — helping tackle some difficult or unexpected problems as we went. Each mode informed the other and propelled further progress. This ended up being one of the most enjoyable periods of my career to date.

A fuller chronicle of the migration project is captured in this slide deck.

Exit

As a result, I started exploring externally, with a focus on Big Tech as I went on parental leave towards the end of 2023.

Takeaways

3.5 years at a startup felt more like 10 years.

My experience at Unit21 represented a stereotypical Bay Area startup journey through an economic cycle. I joined right after its Series A, helped with its rapid growth and two more rounds of relatively easy funding, rode through the sudden downturn and the layoff in 2022, and fought on the perilous road to recovery.

Through the ups and downs, I grew so much as an engineer and as a professional.

People