Machine Learning Engineer at Social-impact AI startup backed by founders of Tide and Mumsnet

You will own the end-to-end detection pipeline for a pioneering child-safety app, developing on-device ML models that protect children from harmful content in real-time. By optimizing YOLO-style detection and transformer classifiers for mobile hardware, you will build a privacy-first safety layer that works across every application on a child's smartphone.

• Design and train computer vision and NLP models to detect harmful content across varied UI contexts and video feeds.
• Optimize model performance for low-latency, on-device execution using quantization, TFLite, and ONNX Runtime on Android devices.
• Build robust evaluation suites and error-analysis workflows to minimize false positives while maintaining high recall for critical safety scenarios.

• 2+ years of experience in applied ML or research, ideally with a background in object detection (YOLO) and transformer architectures (ModernBERT/RoBERTa).
• Proven ability to ship production-grade ML systems, balancing scientific rigor with the practical trade-offs of a fast-moving startup environment.
• Hands-on expertise in model optimization and threshold tuning, specifically for on-device inference where memory and speed are constrained.

• Work on a high-impact product with a genuine social mission, supported by high-profile figures like Sir Tim Berners-Lee and Hugh Grant.
• Own the entire ML capability from strategy to deployment, reporting directly to the founder with significant influence over technical direction.
• Solve complex on-device inference challenges through a unique partnership with Mumsnet and Nothing Technology to launch child-safe hardware.