Keppel·18 days ago
18 days ago
Data Scientist, AI/ML Predictive Maintenance Platform
Sign up free to see how well your resume matches this role.
What you'll do
- Develop, train, and fine-tune models on time-series sensor data for both anomaly detection and failure prediction across various asset types.
- Explore and apply a mix of forecasting, anomaly detection, change point detection, survival analysis, and representation learning techniques, choosing the best fit based on the use case.
- Be comfortable applying hybrid approaches (e.g., combining statistical thresholds with ML models, or chaining changepoint detection with LSTM/Transformer predictors) when appropriate.
- Iterate quickly and deliver working ML models into dev or production environments every 1–2 week sprint, enabling rapid feedback and continuous improvement.
- Balance performance, explainability, compute cost, and deployment constraints in every modeling decision
- Continuously monitor model performance in production (e.g., accuracy, drift, recall).
- Build a retraining and rollback strategy to handle data drift, model drift or edge cases.
- Use dashboards or alerts to track live model degradation.
- Proactively recommend model retirement or replacement.
- Clearly explain model behavior (e.g. thresholds, decision boundaries).
- Share experiment results, performance comparisons, and trade-offs transparently.
- Collaborate with product managers, engineers, and non-technical stakeholders.
Summarised by NextRaise from the employer’s description, which follows in full below.
Full description from employer
JOB DESCRIPTION
We are seeking a skilled and pragmatic Data Scientist to join our team in shaping the next-generation AI/ML-based predictive maintenance platform. You’ll work on applying both classical and Gen AI techniques to real-time IoT signals across diverse physical assets. If you thrive in fast iterations, own your experiments, and believe in getting things to production—not just Jupyter notebooks—this role is
for you.
Key Responsibilities
1. Model Development
- Develop, train, and fine-tune models on time-series sensor data for both anomaly detection and failure prediction across various asset types.
- Explore and apply a mix of forecasting, anomaly detection, change point detection, survival analysis, and representation learning techniques, choosing the best fit based on the use case.
- Be comfortable applying hybrid approaches (e.g., combining statistical thresholds with ML models, or chaining changepoint detection with LSTM/Transformer predictors) when appropriate.
- Iterate quickly and deliver working ML models into dev or production environments every 1–2 week sprint, enabling rapid feedback and continuous improvement.
- Balance performance, explainability, compute cost, and deployment constraints in every modeling decision
2. Model Performance Monitoring
- Continuously monitor model performance in production (e.g., accuracy, drift, recall).
- Build a retraining and rollback strategy to handle data drift, model drift or edge cases.
- Use dashboards or alerts to track live model degradation.
- Proactively recommend model retirement or replacement.
3. Team Player
- Clearly explain model behavior (e.g. thresholds, decision boundaries).
- Share experiment results, performance comparisons, and trade-offs transparently.
- Collaborate with product managers, engineers, and non-technical stakeholders.
- Align ML decisions across the team for unified communication.
- Follow lean documentation principles: be precise, but cover key code, decisions, and the chosen approach
JOB REQUIREMENTS
-
BUSINESS SEGMENT
ConnectivityPLATFORM
Operating DivisionCompany
Keppel
Singapore
Company facts come from this company's own listings. We only show what the postings themselves carry.