A focused ML practice built in Singapore.
We're not a research lab and we're not a large consultancy. We're a small team of ML practitioners who work closely with clients to turn data into working models.
← Back to HomeWhere Nolux came from
Nolux was founded in 2019 by a group of ML engineers who had spent years embedded in large technology organisations across Singapore and the broader APAC region. The pattern they kept encountering was the same: companies with good data and well-framed problems, but no straightforward path to building models that actually worked in their specific context.
The agency model didn't fit. Research teams operated on a different timescale. Freelancers were hard to evaluate. What was missing was a small, accountable team with deep technical skills and a preference for doing fewer things well.
That is what Nolux is. We work with a limited number of clients at a time to give each engagement the focus it needs. Our practice is built around three service types — custom model development, model evaluation, and proof-of-concept builds — because we believe a narrow offering done well serves clients better than a long menu of loosely related work.
What drives our work
Our mission is straightforward: help teams make good decisions about ML by building things that are honest about what they can and cannot do.
Machine learning has a habit of being oversold. Models that look impressive in a demo behave poorly in production. We try to reverse that tendency — working carefully through the data, being clear about limitations, and delivering something that holds up when it reaches real users.
People behind the work
A small group of ML engineers and data practitioners, each with a background in applied research and production systems.
Led model development at a regional fintech before co-founding Nolux. Specialises in classification and recommendation systems.
Background in applied NLP research and data quality engineering. Leads evaluation engagements and model auditing work.
Ensures models reach production with solid data pipelines and deployment infrastructure. Extensive experience with cloud-native ML tooling.
Standards we hold across every engagement
These aren't aspirational values. They reflect how we actually run projects — and they're worth knowing before we begin working together.
Data confidentiality
Every engagement begins with a mutual NDA. Your data is used only for the agreed scope and is not retained after project handover.
Fairness assessment
We test for performance disparities across relevant subgroups before any model is delivered. Fairness isn't a checkbox — it's part of the evaluation process.
Full documentation
All delivered models include model cards, training pipeline documentation, and deployment guidance so your team can maintain and extend the work.
Agreed evaluation criteria
We agree on metrics and acceptance criteria before modelling begins. Success is defined by what matters to your use case, not by default benchmark scores.
Version-controlled code
All code and experiment tracking is managed in version control from the start. You receive the full repository — not just final outputs.
Transparent communication
We communicate clearly when something isn't working. If the data doesn't support the approach, we'll say so — with evidence — rather than delivering something misleading.
Applied machine learning for organisations that value precision
Nolux works with organisations across Singapore and APAC that are navigating practical questions about machine learning: whether a specific problem is suited to an ML approach, how to assess the performance of a model already in production, or how to move from an interesting idea to a working system with minimal risk. Our three service types are deliberately scoped to where these questions typically sit.
Our team brings direct experience with classification, regression, clustering, and recommendation problems across sectors including financial services, logistics, healthcare technology, and B2B SaaS. We work in Python-based ML tooling, maintain clean data pipelines, and deliver production-oriented code rather than research notebooks. For each client, we bring the same focus: understanding the data, agreeing on what success looks like, and building something that earns that designation.
Ready to explore what's possible with your data?
Start with a brief conversation about your problem. No commitment required.
Get in Touch