About
I'm a software engineer in Auckland with ten years of building and running production SaaS. I own backend and integration architecture at two Y Combinator companies: Simbie AI, an agentic voice AI platform for US healthcare, and TimeTackle, a calendar analytics product I've been a founding engineer on since 2021.
The work I like most sits where systems meet: agent harnesses and MCP servers, call routing, CRM and EHR integrations, webhooks that have to be idempotent, and the observability that tells you when any of it has quietly stopped working. Most of it handles patient or enterprise data, so security and privacy are part of the design, not a pass at the end. I write about that work on the blog.
Before New Zealand I spent six and a half years at Brain Station 23 in Dhaka and left as a Staff Software Engineer and technical lead, owning technical direction for US startup clients. Along the way I worked on AXAR AI, an open-source TypeScript agent framework, and taught React and AWS to bootcamp cohorts.
I'm finishing a Master of Applied Technology at Unitec. My thesis asks why voice biometric models fail on populations they weren't trained on, and how to set decision thresholds honestly when they do. Same signal I work with all day, different question.
Open to senior backend, platform and AI engineering roles in New Zealand.
Work Experience
Education
Skills
Things I've shipped
Production voice agents in healthcare, an integration platform and MCP server for a calendar analytics product, and an open-source agent framework. Each card links to a write-up on how it was built.
Voice agents that answer patient calls for US clinics, route them across phone lines and write into the EHR. I built the agent harness, the config-driven multi-agent service and the call-transfer routing behind 187+ production agents, plus the simulation framework that tests agents without touching a real EHR. Isolating heavy Celery tasks took call exports from 25–30 minutes to about one.
Calendar analytics that syncs Google and Outlook with CRMs and practice-management systems. I architected the integration platform behind it — standardised auth, lifecycle and a provider-agnostic data model — built two-way CRM sync that doesn't create duplicates, and shipped an MCP server with MCP-scoped OAuth tokens that's listed in the ChatGPT and Claude connector directories. Moving services to Cloud Run halved the GCP bill.
Open-source TypeScript framework for building production LLM agents — type-first, model-agnostic, with decorators and dependency injection instead of hidden magic. 163 stars. I worked on streamed responses and real-time logging, and built the release pipeline that publishes to both npm and JSR.
An AI sales workspace (YC W21) for people who live in back-to-back calls. I led the build of the cross-platform Electron desktop app: it hooks into Zoom, transcribes each call with Recall AI, and uses GPT-3.5 to analyse what was said and draft the follow-up, so the admin tail of every meeting becomes one click.
A US social banking app that put checking, a line of credit and savings behind one balance and paid cash rewards on spending. I built money transfer with rewards, push notifications and iOS/Android biometric authentication, and wrote the end-to-end tests.
Get in Touch
I'm open to senior backend, platform and AI engineering roles in New Zealand. The fastest way to reach me is email — I read everything and reply to anything that isn't a mass mailout.






