Teams That Have Worked With Us
What engineering teams across Malaysia have found after a session with Halcyon Compute — and what changed in how they run their model deployments.
Back to HomeWhat Teams Say
Reviews from engineers and leads who attended our readiness walkthroughs, tuning workshops, and advisory retainers.
Ahmad Kamal
Platform Engineer · Petaling Jaya
"The readiness walkthrough gave us a concrete list of what to sort out before we pushed the model to production. We had assumed a few things were covered that turned out not to be. The rubric made it easy to hand off to the rest of the team."
May 2025 · Readiness Walkthrough
Nurul Liana
ML Engineering Lead · Kuala Lumpur
"We came to the pipeline tuning workshop with a serving setup that was technically working but felt slow in ways we could not pin down. The two sessions were structured well — we covered batching and caching in a way that actually applied to our setup, not just in theory."
May 2025 · Pipeline Tuning Workshop
Wong Boon Keong
Infrastructure Lead · Shah Alam
"Three months into the retainer, our runbook has grown from a short internal doc into something the whole team actually opens when there is an issue. The scheduled calls kept us honest about the things we said we would sort out."
April 2025 · Advisory Retainer
Selva Rajan
Senior Data Engineer · Cyberjaya
"I appreciated that the facilitators came prepared. They had read through the architecture notes we sent and did not spend the first hour asking things we had already written down. The session time was well-used."
May 2025 · Readiness Walkthrough
Tng Hui Min
MLOps Practitioner · Penang
"The tuning worksheet is genuinely useful. We go back to it when someone on the team questions a configuration decision — the reasoning is documented, not just the outcome. That is rarer than it should be after a workshop."
April 2025 · Pipeline Tuning Workshop
Fauziah Aman
Tech Lead · Johor Bahru
"We started the retainer not entirely sure what value it would add, since we already had our own practices. What changed was that those practices got written down properly. By the end of month two, our runbook had become the reference point for the whole platform team."
May 2025 · Advisory Retainer
How Teams Used Each Engagement
Preparing a Recommender Model for Production
E-commerce platform team · Klang Valley · May 2025
Challenge
The team had trained a recommender model over several months. They were ready to deploy but had not defined what monitoring would look like in production, and the rollback plan existed only as a verbal understanding among two engineers.
What We Did
The half-day walkthrough covered packaging, monitoring setup, and rollback ownership. We completed the readiness rubric as a group, which surfaced three areas — request logging, ownership of the rollback trigger, and drift monitoring — that had not been assigned to anyone.
Outcome
The team assigned the three open items before the session ended. The model went to production two weeks later. No rollbacks were required in the first month, and when latency spiked briefly, the monitoring they had set up identified the cause within minutes.
Reducing Inference Latency on a Document Classification Service
Legal technology team · Kuala Lumpur · April 2025
Challenge
Median inference latency had increased by roughly 40% over three months as request volume grew. The team had made several configuration changes without a clear record of what had been tried and what the results were.
What We Did
Session one reviewed the request path and identified that batching was configured too conservatively for the actual request pattern. Session two worked through revised batching parameters and introduced result caching for a subset of frequently repeated inputs.
Outcome
Median latency dropped by approximately 28% after the configuration changes were applied. The tuning worksheet documented both the changes made and the options that were considered but not implemented, giving the team a basis for further work without repeating the same analysis.
Maturing Operations Across a Four-Model Production Platform
Financial services technology team · Kuala Lumpur · Feb–May 2025
Challenge
The team was running four models in production with different ownership arrangements, different monitoring setups, and no shared documentation. Operational knowledge was distributed across individuals with no single reference point.
What We Did
The three-month retainer started with a baseline review across all four models. We established a shared runbook structure in month one, held monthly review calls, and updated the runbook after each call to reflect decisions made and changes applied.
Outcome
By the end of month three, all four models had consistent runbook entries. A new engineer who joined the team mid-engagement was able to orient themselves using the runbook without requiring one-to-one walkthroughs from the rest of the team.
By the Numbers
80+
Teams supported
4.7/5
Average session rating
6
Years of advisory practice
3
Business days to written output
Get in Touch
Phone
+60 3-8312 6904Address
No. 8, Persiaran APEC, Cyber 8, Cyberjaya
Hours
Mon–Fri: 9AM–6PM
Ready to Have Your Own Session?
Contact us to discuss which engagement fits your team's current situation. We respond within one business day.
Book a Session