Machine Learning Engineer at SafetyMode — AI-powered child safety startup

You will own the entire detection pipeline for the world's most advanced on-device child safety technology. From dataset strategy and model training to production deployment, you will build real-time AI that identifies harmful content across all smartphone apps. This role offers the unique challenge of balancing high-performance computer vision with strict on-device privacy.

• Design and train robust object detection and transformer-based text classification models to identify various categories of online harm.
• Optimize model architectures for low-latency execution on Android devices using quantization, TFLite, and novel on-device inference strategies.
• Establish rigorous evaluation suites and automated labeling pipelines to continuously reduce false positives in complex, high-noise mobile environments.

• At least 2 years of experience in applied ML with a track record of shipping production-grade computer vision or NLP systems.
• Proficiency in modern detection pipelines (YOLO) and transformer architectures (RoBERTa/ModernBERT), with a deep understanding of precision/recall tuning.
• Hands-on experience with on-device optimization tools like ONNX Runtime or Core ML and a passion for solving ambiguous, high-stakes safety problems.

• Rare opportunity to own the end-to-end ML capability for a product with a massive social mission and high-profile backing from Sir Tim Berners-Lee and Hugh Grant.
• Work on cutting-edge technical challenges involving real-time, on-device inference (YOLO, Transformers, TFLite) where your code directly protects thousands of children from harm.
• High-impact founding team environment incubated by CanDo, with strategic partnerships including Mumsnet and Nothing Technology to scale globally.