I work on the distance between a commit and production.

Route in

The route in was an apprenticeship, not a graduate scheme. Adex International took me on as an Apprentice Solutions Architect in November 2024, where an AWS design had to be defended on availability, cost and blast radius before anything was provisioned, and where the answer to “can we just” was usually a diagram of what would break.

Fusemachines followed in May 2025 with a fellowship in AI and MLOps engineering. That year was spent on containerised training pipelines, inference services on K3s, and the realisation that a model service has one failure mode the rest of an estate does not: it can be healthy, fast and wrong at the same time. Request metrics alone will never tell you that, which is why prediction distribution belongs on the same dashboard as latency.

In March 2026 acAIberry brought me in as a Trainee DevOps Engineer on a healthcare platform operating under HITRUST. Regulation changes what a pipeline is for. A stage stops being convenience automation and becomes an audited control, which means it has to leave evidence behind rather than a log line that rotates away in a fortnight.

Study alongside the work

The degree runs underneath all of it: BSc (Hons) Computer Science with Information Technology at Lord Buddha Education Foundation, in partnership with Asia Pacific University, from 2023 to 2027. Reading about a distributed system and being on call for one are different activities, and doing both at once is the reason each half is useful to the other.

What I hold to

Infrastructure should be described rather than assembled. Terraform is the primary artefact, not documentation written after the fact, and a module that cannot be read in a pull request is not reviewable no matter how well it plans. Deployments are identified by image digest rather than a mutable tag, because a tag that moved is an outage with no audit trail. A production gate that cannot say no is decoration.

Observability is a design activity, not an afterthought bolted on once something has already fallen over. The useful question is not whether metrics exist but whether they answer the question you will actually ask at two in the morning, and that is a decision made while the system is being built.

Teaching and community

The other half of the work is community. I am an AWS Community Builder, I co-founded the AWS Student Builder Group at Lord Buddha Education Foundation, and I served as College Representative for AWS Cloud Club Nepal for the 2025/26 tenure. Most of that time goes on teaching: Git and GitHub for people whose entire experience of version control is a download button, and AWS fundamentals for people who have been told the cloud is simple. Hands on keyboards, not slides.

Online, and on every repository behind this site, I am gocools.