Data Scientist, AI/ML Predictive Maintenance Platform
Singapore Full-time Mid-level AI Research
At a glance
Mid-level AI Research role at Keppel. Singapore · full-time.
Pay not stated AI Research salary
RoleAI Research
SeniorityMid-level
LocationSingapore
EmploymentFull-time
PostedSep 2, 2026 · 7w ago
Role brief
aimlroles summary, based on the employer's posting.
What you'll do
- Build and test sensor-data models that forecast asset health, flag anomalies, predict failures and estimate time to failure
- Compare modelling methods against measurable criteria, balancing detection, false alarms, explainability and deployment needs
- Design experiments for sparse failure records, using approaches such as weak supervision, proxy labels and transfer learning
- Put models into development and production, track their performance and data quality, then set retraining and rollback rules
- Turn operational issues into machine-learning work with product, engineering and domain experts, including alert interpretation
What you bring
- A degree in engineering, computer science, data science or a related area; AI/ML study is preferred
- Four to seven years in data science or applied machine learning in software or engineering settings
- At least two to three years building failure prediction, anomaly detection, condition monitoring or maintenance models from sensor data
- Working knowledge of time-series methods, anomaly detection, classification, regression and model evaluation
Who this fits
This mid-level role suits someone who combines applied modelling with careful experimentation, especially when failure examples are scarce. You’ll work with product, software and platform teams as well as mechanical and electrical domain experts.
From the employer
JOB DESCRIPTION
We are looking for a Data Scientist to build and operationalize machine learning solutions for failure prediction and early-warning detection across data center and industrial assets.
The role focuses on identifying degradation patterns, forecasting asset health, detecting anomalies, and estimating time to failure from sensor data. Success requires strong experimentation skills, the ability to work with limited failure examples, and a disciplined approach to validating models in real-world environments.
Candidates should demonstrate both technical depth and a strong research mindset.
Key Responsibilities
1. Predictive Maintenance & Early-Warning Model Development
- Develop and validate models using operational sensor data for failure prediction, asset (mechanical and electrical equipment) degradation monitoring, early-warning detection, remaining useful life estimation, and time-to-threshold forecasting.
- Apply and compare survival analysis, time-series forecasting, anomaly detection, change-point detection, and representation learning techniques.
- Select modelling approaches using measurable performance criteria and documented experimental results.
- Balance detection performance, false-positive rates, explainability, and production deployment requirements.
2. Working with Sparse Failure Data
- Develop strategies for environments where no failure events.
- Work with limited labelled failures, class imbalance, weak supervision, proxy labels, transfer learning, and failure-data sourcing.
- Establish validation methods that accurately measure model effectiveness despite limited failure examples.
3. Experimentation & Research
- Form hypotheses and design experiments to evaluate competing approaches.
- Define baseline approaches, evaluate alternatives, measure improvements using objective metrics, and document findings and limitations.
- Justify model selection using evidence rather than preference.
4. Model Productionization & Monitoring
- Deploy models into development and production environments.
- Monitor precision, recall, drift, and data quality.
- Establish retraining, rollback, and retirement criteria for production models.
- Investigate performance degradation and implement corrective actions.
5. Cross-Functional Collaboration
- Collaborate with product managers, software engineers, platform engineers, and subject matter experts (data center domain with mechanical and electrical engineering expertise)
- Translate operational problems into machine learning problems.
- Incorporate domain expertise into model development, validation, and alert interpretation.
JOB REQUIREMENTS
Requirements
- Bachelor’s or Master’s degree in Engineering, Computer Science, Data Science, or a related field; specialization in AI/ML is preferred. Mechanical engineering background or strong exposure to mechanical/industrial systems will be an advantage.
- 4–7 years of relevant experience in data science, or applied machine learning, either in a large software development organization or an end-user engineering/industrial environment.
- Minimum 2–3 years of hands-on experience developing failure prediction, anomaly detection, condition monitoring, or predictive maintenance models using engineering, operational, or IoT sensor data.
- Good understanding of common machine learning approaches and the intuition behind them, including time-series models, anomaly detection methods, classification/regression models, and model evaluation techniques.