40+
Completed engagements
4.8
Average client rating
92%
Clients who return or refer
6
Years in the market
In their own words
"We had about 18 months of recorded call centre audio sitting in storage, completely unusable. Within 10 weeks, omnifoldaxa had built a pipeline that processed it all and surfaced topics we had never been able to see before. The documentation was thorough enough that our in-house team could take over without any hand-holding."
Nurul Rashidah
Head of Operations, Klang Valley call centre
March 2025
"We wanted to test whether a visual inspection system could work for our production line before committing significant budget. The computer vision prototype omnifoldaxa delivered was exactly that — a working demo we could show stakeholders. It ran on time, within the quoted price, and the evaluation report made the business case easy to present."
Tan Li Ming
Engineering Manager, Selangor manufacturing firm
February 2025
"Our internal search tool had become a running joke among staff. After omnifoldaxa's relevance tuning engagement, the improvement was noticeable within a week of the new ranking model going live. The team was communicative throughout — we always knew what stage we were at and what was coming next."
Syafiq Baharuddin
Technology Lead, KL financial services group
January 2025
"We brought omnifoldaxa in to look at our interview transcription workflow. What I appreciated was that they were honest early on about what was achievable with our existing data. That kind of candour is rare. The system they delivered handles Manglish and code-switching far better than anything we'd tried before."
Priya Vasudeva
Research Director, PJ market research agency
March 2025
"We run an e-commerce platform with around 80,000 SKUs. Search quality was causing measurable cart abandonment. After omnifoldaxa's search relevance engagement, we saw a notable improvement in search-to-purchase conversion. The scope of work document they sent at the start was very clear — there were no misunderstandings about what was included."
Chen Kai Weng
Product Manager, Shah Alam e-commerce platform
February 2025
"Our legal team had been manually reviewing depositions for years. The speech pipeline omnifoldaxa built changed that significantly. The speaker diarisation accuracy was good enough for our use case from the first release. Onward maintenance has been straightforward — the documentation made it manageable for our in-house developers."
Ahmad Mustaqim
Managing Partner, KL legal practice
January 2025
Project stories
From recorded backlog to searchable knowledge base
Challenge
A KL call centre had three years of customer call recordings in storage — roughly 200,000 hours — with no practical way to analyse them. Compliance requests took weeks of manual review.
Solution
We built an end-to-end speech-to-insight pipeline with custom acoustic model fine-tuning for the client's call centre vocabulary, speaker diarisation, and a searchable output format integrated with their existing compliance dashboard.
Outcome
Compliance review time dropped from weeks to hours. The backlog was processed in full within two weeks of system deployment. Team: 3 people. Timeline: 11 weeks. Delivered within scope.
Validating shelf-counting before full investment
Challenge
A retail chain wanted to automate in-store shelf counting but was unwilling to commit to a large development contract without first validating that the approach was technically feasible.
Solution
We ran a five-week computer vision prototyping engagement. This included building a small labelled dataset from store images, training a detection model, and deploying a functional demo on a staging environment with a performance report.
Outcome
The prototype achieved 88% detection accuracy on the test set — above the client's minimum threshold. The client proceeded to a full production build with a separate team, using our prototype and evaluation report as the technical foundation.
Improving article discovery for a news archive
Challenge
A Malaysian digital publisher had a 15-year article archive with a keyword search system that struggled with queries in Bahasa Malaysia, English, and mixed-language terms common among local readers.
Solution
We layered a learning-to-rank model over their existing search engine. The process included query log analysis, relevance judgement collection from editorial staff, multilingual feature engineering, and model training with NDCG-based offline evaluation.
Outcome
NDCG@10 improved by 34% on the offline evaluation set. Reader engagement time on search result pages increased measurably following deployment. Timeline: 7 weeks. No changes to existing search infrastructure required.
Reach us directly
Phone
+60 3-2143 7862Address
21 Jalan Bukit Bintang, KL
Hours
Mon–Fri 9am–6pm
Sat 10am–2pm
Professional affiliations and standards
SSM Registered
Companies Commission of Malaysia
PIKOM Member
Association of ICT industry Malaysia
PDPA Compliant
Personal Data Protection Act 2010
MSC Malaysia Status
Since 2021