The Team Behind the Runbook
Halcyon Compute was built around one idea: that operational clarity in model deployment comes from structure, not from experience alone.
Back to HomeGrounded in How Teams Actually Work
Halcyon Compute started from a straightforward observation: many engineering teams building capable models were underserved when it came to deploying and maintaining them. The tooling landscape had expanded, but the operational knowledge — the habits, checklists, and decision-making frameworks — lagged behind.
We set up in Cyberjaya to be close to the engineering community that drives much of Malaysia's technology sector. Our work began with small teams preparing their first production deployments and grew to include larger platforms managing multiple models. Each engagement sharpened our understanding of what teams need versus what they already have.
Today, Halcyon Compute offers three structured engagements — a readiness walkthrough, a pipeline tuning workshop, and a multi-month advisory retainer — each designed around tangible outputs that teams keep and use. We do not sell platforms or tooling. We help people think more clearly about the operational side of what they have built.
Dependable Practices, Not Just Advice
Clarity Over Complexity
Every session produces documentation your team can open on a difficult deployment day. We write for the person who needs to act, not the person who commissioned the report.
Team-First Engagement
Our sessions are facilitated collaboratively. The goal is for your team to finish with a shared vocabulary and shared ownership of the practices we discuss — not just a list of recommendations handed over.
Vendor-Neutral Standpoint
We advise on principles and operational habits. We do not have preferred tooling vendors and do not benefit from steering you toward any particular platform. What we recommend is what we judge to be right for your setup.
Our Team
Three practitioners with backgrounds in platform engineering, MLOps advisory, and technical writing.
Razlan Hakim
Principal Advisor, Deployment Operations
Razlan has spent the past eight years working with platform engineering teams across Southeast Asia. He leads readiness walkthroughs and oversees the retainer programme.
Suraya Noor
Workshop Facilitator, Serving Pipelines
Suraya focuses on inference serving and resource scheduling. She designs and facilitates the pipeline tuning workshops, translating abstract optimisation concepts into session-ready exercises.
Farid Tarmizi
Operations Documentation Lead
Farid is responsible for the written outputs that come out of every engagement — checklists, worksheets, and living runbooks. He has a background in technical writing and production documentation for distributed systems.
How We Work
The practices we apply to our own engagements reflect the operational disciplines we advise on.
Pre-Session Preparation
Before each engagement, we review any context the team shares — architecture diagrams, deployment notes, or recent incident summaries — so session time is spent on decisions, not introductions.
Confidentiality as Standard
All information shared with us during engagements is treated as confidential by default. We do not reference client setups in other engagements and do not share operational details beyond what the client explicitly permits.
Documented Outputs Always
Every engagement produces at minimum one written output — a rubric, worksheet, or summary — delivered within three business days. Retainer clients receive a maintained runbook throughout the engagement period.
Review and Feedback Loop
After each session, we send a brief summary of what was covered and any open items. Teams can respond with corrections or additional context. This keeps the written output accurate to what the team actually decided.
Grounded Recommendations
We do not recommend practices we would not apply ourselves. If a suggested approach carries trade-offs, we name them plainly. Our aim is to leave teams better-informed, not more committed to a particular path.
Continuous Practice Development
The operational landscape for model serving changes regularly. We update our session materials and checklists in response to shifts in tooling patterns, infrastructure norms, and the practical issues teams report across our engagements.
Deployment Operations as a Discipline
Model deployment has matured considerably over the past several years. Serving frameworks, orchestration platforms, and monitoring tooling have each developed into areas of real specialisation. What has kept pace less neatly is the operational knowledge that sits around those tools — the question of who reviews a deployment before it goes live, how rollback decisions get made under pressure, and what documentation a team needs to hand off a serving pipeline to another group.
Halcyon Compute works in that space. Our engagements are practical and scoped: a half-day to assess readiness before a launch, a two-session workshop to review and improve a serving setup, or a three-month arrangement for teams that want sustained operational support across several models. We focus on the habits, structures, and written records that make production ML operations reproducible and understandable over time.
Our team draws on experience across batch inference, real-time serving, and scheduled prediction workloads. We have worked with teams building on managed cloud platforms, on-premises Kubernetes clusters, and hybrid arrangements. The common thread is an emphasis on operational clarity — clear ownership, documented rollback paths, and monitoring that signals something actionable rather than simply logging that something happened.
Based in Cyberjaya, we are positioned to work with teams across Malaysia's technology corridor, from Klang Valley to Penang. Remote engagements are fully supported for teams elsewhere in the region.
Start with a Conversation
Tell us where your team is in its deployment journey and we will suggest the most appropriate starting point. No obligation, no sales process.
Contact Halcyon Compute