From the Organisations
We've Worked With
Honest accounts from clients across healthcare, financial services, and enterprise technology.
← Back to HomeWhat Clients Say
"We'd been sitting on two years of chatbot logs and had no real insight into what our customers were actually asking. The analytics layer Cortexium built gave us a clear picture within weeks. The dashboard is something our team actually uses in weekly planning."
"Data privacy was a hard constraint — we couldn't share patient records across our facilities for any kind of centralised training. The federated approach Cortexium designed worked within those constraints without needing to change our data governance policies."
"I wanted to know if our idea was technically feasible before committing serious budget. The PoC Sprint gave us a working prototype and an honest answer — the recommendation was to adjust the problem framing, which actually saved us from a costly misdirection."
"The scoping process was thorough — they asked questions I hadn't thought to ask about our data pipelines. By the time we started development there were very few surprises, which made the timeline reliable. Weekly updates were brief but actually informative."
"We used the PoC Sprint to test a document classification approach before committing to a larger architecture investment. The prototype ran against our actual data. The write-up was technical enough to share with our board as a credible assessment."
"The federated learning engagement addressed a coordination challenge we'd been deferring for over a year because no one had a clear answer on how to do it. Cortexium came with a structured architecture proposal in the first two weeks. The final handover documentation was genuinely thorough."
Engagements in Detail
The Challenge
A regional insurance provider had deployed a WhatsApp-based claims assistant. Conversation volumes were high but the team had no view of which queries were being resolved versus deflected, or where users were dropping off.
The Approach
We extracted and processed 14 months of conversation logs, applied an intent taxonomy, and built a resolution funnel model. A Looker Studio dashboard was configured to surface daily and weekly patterns for the operations team.
The Outcome
The team identified three high-frequency intent clusters that were generating unnecessary escalations. Adjustments to the assistant's handling of those clusters reduced escalation volume by approximately 28% over the following quarter.
The Challenge
A network of five clinics across Selangor wanted to train a shared patient triage model. Each clinic managed its own patient data under separate PDPA obligations — centralising records for training was not viable.
The Approach
We designed a federated training architecture with a central aggregation server and local training nodes at each clinic. Secure aggregation was implemented so that model updates — not patient data — were shared. Training ran across 12 weeks with regular convergence checks.
The Outcome
The federated model matched the performance of a centralised baseline on held-out test data. Each clinic retained full ownership of its data. The architecture documentation was submitted as part of the network's internal compliance review.
The Challenge
A logistics technology startup believed that their delivery exception data could be used to predict future failure points before they occurred. They needed to know whether their two years of data were sufficient and whether the hypothesis was technically sound.
The Approach
We scoped a binary classification prototype, ran data quality checks, and built a baseline model against the available historical records. Evaluation was conducted against a held-out period to simulate real-world prediction conditions.
The Outcome
The prototype showed moderate predictive capability, with recommendations to expand the feature set before full development. The go/no-go document outlined the data gaps and the specific collection steps needed to make a production model viable.
Talk to Our Team
Mon–Fri 9:00 AM – 6:00 PM
Sat 9:00 AM – 1:00 PM
Add Your Own Experience
Start with a short conversation about your situation. We'll tell you which service makes sense — or whether the timing isn't right.
Get in Touch