Senior Machine Learning Engineer | Dubai Or Bangalore Based.
Summary
Build and tune adaptive ML models (XGBoost, Isolation Forests, autoencoders) to score OT security alerts, classify industrial devices, and detect threats on air-gapped appliances using self-hosted MLOps.
Job description
Machine Learning Engineer – Classical & Adaptive ML
Discover the Opportunity
We are hiring a hands‑on Machine Learning Engineer to own the classical ML powering Forge’s detection and triage layer. Forge protects critical Operational Technology (OT) and industrial control environments.
In this individual contributor role, you will build and tune models that score and prioritize security alerts, classify OT assets, and separate real threats from noise—all running directly on local, air‑gapped appliances.
Because every deployment operates in a unique, isolated OT environment without access to cloud MLOps services, our models can't be static artifacts shipped once. You will build adaptive machine learning systems that learn on the edge, evolving over time based on local traffic patterns and analyst feedback.
Discover the Role
Alert Triage & Scoring: Build and tune XGBoost and gradient boosting models to rank security alerts, drastically cutting false positives and surfacing genuine incidents for analysts.
Adaptive & Online Learning: Design models that adapt to individual site environments over time using incremental learning, concept drift detection, and human-in-the-loop feedback.
Asset & Device Classification: Expand our Equipment Type Inference capabilities using tree ensembles and clustering to profile industrial devices from raw network fingerprint data.
Custom Anomaly Detection: Implement Isolation Forests, one‑class methods, and autoencoders alongside our OpenSearch layer to catch threats that static rules miss.
Risk Scoring: Refine features and models to ensure Forge’s risk scoring accurately reflects real operational exposure.
Production Pipelines: Build reproducible, version‑controlled feature engineering, training, and evaluation pipelines that move beyond standard notebooks.
Self-Hosted MLOps: Stand up and run a self-hosted MLflow setup for experiment tracking, model registry, and promotion of vetted models to production.
On‑Device Model Lifecycle: Package models to run seamlessly within the appliance Docker stack, complete with evaluation gates, drift monitoring, and controlled on‑device retraining.
Discover the Requirements
4+ years of hands‑on experience building and shipping machine learning systems in production.
Gradient Boosting Expertise: Strong Python fluency with deep experience in XGBoost, LightGBM, or CatBoost.
Classical ML & Anomaly Toolkit: Solid grounding in scikit‑learn, tree ensembles, clustering, anomaly detection (Isolation Forests, autoencoders), and selective deep learning.
Adaptive ML Experience: Practical experience with online/incremental learning, concept drift, and active learning/feedback loops.
ML Pipelines & MLOps: Proficiency with pipeline orchestrators (Airflow, Prefect, Dagster, or Kubeflow), data/model versioning (DVC), and model tracking (MLflow).
Imbalanced Data Expertise: Deep familiarity with rare‑event metrics (Precision, Recall, PR curves, calibration) rather than raw accuracy.
Telemetry Data: Comfort working with large volumes of structured logs and network telemetry.
Discover the Desired
Exposure to OT,ICS, or SCADA security, industrial protocols , or the Purdue model.
Experience with OpenSearch or ES machine learning capabilities.
Background deploying ...