Machine Learning Engineer

* The Machine Learning Engineer reports to the Group Service Support Specialist and takes ownership of the key function of Data Science and ML engineering, * This role primary function is to turn raw fleet telemetry, scheduling inputs and operational data into decisions, predictions, and insight that improve the performance of mining operations. * This is the foundational analytical hire — responsible for exploring data, defining problems, collaborating with domain experts, and building models that inform the product roadmap across predictive maintenance, anomaly detection, operator behaviour, and dispatch optimisation. * Typical features to commercialise are multivariate anomaly detection, failure prediction, estimate remaining life prediction, schedule optimisation, OR(Vehicle routing problem), Computer vision solutions. * The role is expected to be able to take ownership of the entire chain of commercialised ML features in the product, including formulating problems, identifying and implementing data infrastructure requirements and implementing ML operational models.

Job Overview:

Data exploration and problem formulation

• Conduct structured data audits on incoming data streams, assessing signal quality, completeness, historical depth, sampling frequency, and suitability for each modelling use case
• Translate ambiguous business questions into scoped problem statements defining target variable, features, data requirements, modelling approach, and success criteria
• Distinguish between problems requiring machine learning, statistical analysis, or rules-based logic and recommend the right approach for each.
• Maintain a prioritised problem inventory with documented assessments of data readiness, tractability, effort, and business value.
• Identify data inventory deficiencies and formulate plans to solve and or retrieve.

Data Architecture and Engineering

• Design and implement ingestion pipelines
• Build and maintain ETL pipelines transforming raw sensor streams into clean, structured, model-ready feature sets
• Design data warehousing and architecture that is fit for purpose.
• Monitor pipeline health, data freshness, and schema drift as the fleet and product evolve.
• Vender selection and justification with respect to the expected outputs and associated costs

Model Development and Validation

• Maintain a structured model registry tracking all experiments, versions, hyperparameters, and evaluation metrics to ensure full reproducibility of any model at any point in time.
• Implement monitoring frameworks for deployed models that track prediction quality, input data distributions, and output stability over time.
• Define and document retraining protocols for each deployed model specifying trigger conditions, data requirements, and the validation criteria a model must pass before replacing the current version.
• Collaborate with engineering to package validated models into production-ready artefacts with clearly documented input formats, output schemas, and failure handling expectations
• Evaluate and recommend appropriate deployment patterns for each model based on latency, reliability, and operational requirements.

ML Ops and production readiness

• Validated and commercially viable models to be developed into production ready solutions for the product.
• Track all experiments, model versions, hyperparameters, and evaluation metrics in a structured model registry ensuring full reproducibility
• Implement monitoring frameworks for deployed models tracking prediction quality, input distributions, and output drift over time
• Establish automated alerting for model degradation and define retraining triggers based on performance thresholds
• Design retraining protocols specifying trigger conditions, training data windows, and validation gates a model must pass before replacing the current version
• Collaborate with engineering to package validated models into production-ready artefacts with clear input formats, output schemas, and failure handling specifications
• Evaluate and recommend appropriate serving infrastructure for each model — batch, near-real-time API, or edge deployment — based on operational requirements

Communication Insights and stakeholder engagement

• Present model findings and analytical results to non-technical leadership in plain language with honest confidence estimates and clear business implications.
• Communicate proactively when data assumptions underpinning a modelling initiative are not met, recommending corrective action rather than proceeding on a weak foundation.
• Document all analytical decisions, model assumptions, and data limitations in a form accessible to engineering, product, and future data science hires.

Skills, Knowledge & Attributes Required:

• Bachelor's degree in engineering, mathematics, computer science, data science, or statistics.
• Postgraduate qualification in data science, machine learning, or applied mathematics is strongly preferred.
• Minimum 5 years in an applied data science or analytical role.
• Strong foundations in statistics, probability, and applied mathematics.
• Proven experience with industrial IoT or operational sensor data — time series analysis, signal quality assessment, anomaly detection, and predictive modelling.
• Proficient in Python and SQL; experienced with Kafka or MQTT, cloud platforms, Snowflake, Databricks, and Git.
• Solid understanding of MLOps practices including experiment tracking, model versioning, monitoring, and deployment.
• Experience with linear or mixed-integer programming is advantageous.
• Exposure to LLM, NLP, or agentic AI is advantageous but not required.