What Our Clients Say
Honest feedback from organisations we've had the opportunity to work with across Malaysia.
Back to HomeClient Reviews
Nurul Wahida
VP Operations, Fintech Startup · Petaling JayaWe engaged Pragmix for a recommendation engine on our lending platform. The team took time to understand how our users actually behave — not just what the data showed on the surface. The engine they built has noticeably improved how we match borrowers with appropriate products. Very pleased with the collaboration.
24 January 2026Cheng Tao
CTO, Digital Media Company · Kuala LumpurWe needed audio analytics across our podcast library — multilingual content in BM, English, and Mandarin. Pragmix handled the complexity well. The pipeline works reliably and the dashboards are actually useful for our editorial team. One minor delay in delivery, but the team communicated proactively.
2 February 2026Priya Subramaniam
Head of Compliance, Regional Bank · KLThe explainability service was exactly what we needed for our credit scoring model. Pragmix produced clear, regulator-friendly documentation and a dashboard our compliance team can actually use. The model cards are thorough and have already been cited in our internal audit reports.
30 January 2026Jason Lim
Product Manager, E-commerce Platform · PenangGood experience with the recommendation system build. The A/B testing framework they included has been genuinely useful — it lets us compare algorithm variations on live traffic. Would have liked a slightly faster turnaround on the initial scoping phase, but the final product is solid.
18 January 2026Aisyah Zain
Director, Healthcare Analytics Firm · Shah AlamWe appreciated how carefully Pragmix handled our patient interaction audio data. Their PDPA awareness was strong, and the resulting analytics pipeline has given our QA team new insight into call patterns. The documentation was particularly well-prepared.
8 February 2026Raj Mohan
AI Lead, Insurance Company · Kuala LumpurWe brought Pragmix in for explainability on our claims processing model. What impressed me was how they translated technical SHAP analysis into language our non-technical board members could follow. That bridge between technical and business understanding is rare to find.
5 February 2026Success Stories
Helping a Penang E-Commerce Platform Surface Better Product Suggestions
Challenge
The platform's existing recommendation logic was rule-based and struggled with their growing product catalogue of over 15,000 SKUs. User engagement on suggested products had plateaued.
Solution
We built a hybrid recommendation engine combining collaborative filtering with content-based methods, trained on 18 months of user behaviour data. The system included real-time and batch processing, plus built-in A/B testing.
Results
Within three months of deployment, click-through rates on recommended products increased by 34%. The average session duration also showed improvement. The project was completed in ten weeks.
"The A/B testing framework alone was worth the engagement. We can now measure what actually works instead of guessing." — Jason L., Product Manager
Making Credit Scoring Transparent for a Regional Bank
Challenge
The bank's ML-based credit scoring model was performing well technically, but internal auditors and regulators needed better visibility into how individual decisions were being made.
Solution
We applied SHAP and feature importance analysis to the existing model, built an interpretation dashboard, and created per-decision explanation APIs. We also produced a compliance-ready model card for regulatory review.
Results
The bank passed its internal audit review with the new documentation. Compliance team can now independently assess any individual scoring decision. The entire engagement was completed in five weeks.
"For the first time, our compliance team can explain an AI decision to a customer in plain language." — Priya S., Head of Compliance
Unlocking Insight from a Multilingual Podcast Library
Challenge
A digital media company had over 2,000 podcast episodes across three languages but no way to search, categorise, or analyse the audio content at scale.
Solution
We built a multilingual audio processing pipeline with transcription, speaker diarisation, topic extraction, and keyword monitoring. The output was fed into an editorial dashboard tailored for content planning.
Results
The editorial team can now search across their full audio library by topic, speaker, or keyword. Content planning cycles shortened, and they discovered listener engagement patterns that informed new series development.
"We went from having a massive archive we couldn't search to having genuinely useful insight about our own content." — Cheng T., CTO
Our Track Record
Contact Information
Sat: 10AM–2PM
Want to Be Our Next Success Story?
We'd be glad to hear about your project and discuss how we might help. Every engagement starts with a simple conversation.
Start a Conversation