Work

At Regie.ai, I lead the Integrations team. We build the systems that connect Salesforce and HubSpot to Regie, bring customer data in and keep changes moving in both directions. I design the full lifecycle of those connections — from setup and initial import to continuous sync, outbound updates, recovery and the visibility people need to trust what happened.

The hard part is that no two customer CRMs mean quite the same thing. Schemas, permissions and sales processes differ, provider APIs change, and long-running work can stop halfway. I design for that reality: keep workspaces isolated, make progress durable, recover safely and verify that data reached the system it was meant for.

The work is not only technical. I work closely with customers to understand the sales process behind the data. With Commvault, one of Regie’s largest customers, I partnered with their leadership from onboarding through renewal.

We mapped how leads entered their funnel, how ownership and stages changed, where automation handed work to people and what their teams expected from both Salesforce and Regie. That understanding helped us shape the product around how they actually worked, support adoption and earn their renewal.

At Nymble, I designed the software between a recipe and the cooking robot. I was responsible for the high- and low-level design of the machine interaction logic: turning composite recipes into macro, micro and liquid dispenses alongside cooking instructions with quantities, timing and ordering the hardware could honour.

The responsibility extended through execution and verification. Parallel dispenses could not collide, “no dispense” states had to be explicit, and failures needed to be reconstructable. I designed the instruction logs and a test bench that compared what the machine did with what it had been told.

Working this close to hardware meant correctness could not stop at an internally consistent output. The design had to reflect physical constraints and make the machine’s behaviour understandable when software and reality did not agree.

Incubyte was my first full-time engineering role. As a Software Craftsperson, I worked full-stack for Republic Finance, a US consumer lender, across services in Java with Micronaut and Go with Gin.

My responsibility was not only to ship changes but to leave the system safe for the next change. We made the behaviour explicit, wrote the test that described it, implemented the smallest code that passed and then cleaned it up. Pairing, review and code that a colleague could read cold were everyday expectations.

That role made quality a daily team practice rather than a final gate. I still work the same way: make the behaviour clear before implementation, prefer the simplest design that holds and treat readability as a shared responsibility.

At Siemens, I researched whether reinforcement learning could control heating and cooling while keeping rooms within ASHRAE comfort bounds and reducing energy use.

The role combined research with engineering. I built the controller and the system around it: containerising models, enabling remote execution and validating the models that already existed. The work later became a challenge at the Siemens Global Hackathon 2020.