ML Engineer — Predictive Maintenance Lahore, Pakistan / Remote ·

OpsSense Technologies Inc.; Lahore; Remote; PKR 2,790,000 - 4,464,000; Full time

Job summary

OpsSense Technologies Inc. is seeking an ML engineer to build predictive maintenance models that forecast bearing failures using multivariate sensor data. You will deploy models that maintenance planners can trust, prioritizing reliability over leaderboard metrics.

You will collaborate with firmware and backend teams to extract features at the edge and in the cloud, design rigorous evaluation, and integrate outputs into operator workflows with clear explanations for non-ML colleagues.

Qualifications

• 3+ years deploying ML in production on time-series or sensor data.
• Strong Python and standard ML stack, not just deep learning.
• Familiarity with signal processing (FFT) or quick to learn it.
• Rigorous evaluation with calibrated models and clear failure modes.
• Able to explain model behavior to non-ML colleagues.

Responsibilities

• Develop and productionize failure-prediction and anomaly-detection models on multivariate sensor time series.
• Build training and evaluation pipelines for noisy, imbalanced industrial data.
• Collaborate with firmware and backend teams on edge vs cloud feature extraction.
• Define evaluation metrics: precision-recall, lead-time, calibration for operations.
• Integrate model outputs into agent workflows with human-in-the-loop explanations.

Skills

Time-series ML

Anomaly detection

Python

Signal processing (FFT)

Model explainability

Tools

SciPy / NumPy

Job description

You will build the models that tell a plant a bearing will fail before it does. The inputs are vibration, temperature, current, and work-order history from our own sensors; the output is a prediction a maintenance planner can trust. Accuracy here is not a leaderboard number — it is avoided downtime.

• Develop and productionize failure-prediction and anomaly-detection models on multivariate sensor time series.
• Build training and evaluation pipelines that handle noisy, imbalanced, real-world industrial data.
• Work with firmware and backend teams on feature extraction at the edge versus the cloud.
• Define honest evaluation: precision-recall trade-offs, lead-time metrics, and calibration that operations teams can act on.
• Feed model outputs into agent workflows with human-in-the-loop controls and clear explanations.

What you bring

• 3+ years applying machine learning in production, ideally on time-series or sensor data.
• Strong Python and experience with the standard ML stack; solid grasp of classical methods, not just deep learning.
• Understanding of signal processing (FFT, spectral features) or willingness to learn it quickly.
• Rigor about evaluation and failure modes; you would rather ship a calibrated simple model than an impressive uncalibrated one.
• Ability to explain model behavior to non-ML colleagues and customers.