Machine Learning Engineer

Role Overview

Machine Learning Engineer (3–5+ years) at 10a Labs working across the full ML lifecycle to build production-ready systems for AI safety, security, and intelligence use cases.

Responsibilities

• Design, train, evaluate, and deploy ML models across text, image, audio, and multimodal domains.
• Develop and improve classification systems for safety, security, abuse detection, and intelligence.
• Run experiments to benchmark and compare AI models, including LLMs and multimodal systems.
• Contribute to model distillation, optimization, and fine-tuning for performance and deployability.
• Design evaluation pipelines, metrics, and testing frameworks to measure capabilities, reliability, and safety.
• Build agentic systems and automated workflows for evaluation, red teaming, research, and large-scale experimentation.
• Own projects end-to-end from research/prototyping through deployment and monitoring.
• Partner with software engineers to productionize systems and support ongoing improvements.
• Provide technical expertise during client engagements and internal initiatives.

Requirements

• 3–5+ years professional experience building and deploying ML systems.
• Strong Python skills and proficiency with PyTorch and/or TensorFlow.
• Experience across multiple modalities, with expertise in one or more of:
  • Computer Vision (e.g., classification, detection, OCR, segmentation, deepfake detection, multimodal vision-language)
  • NLP (e.g., LLMs, text classification, information extraction, retrieval, speech-to-text, agentic applications)
• Experience training, fine-tuning, evaluating, and deploying ML models in production.
• Experience designing evaluation methodologies, benchmarking systems, and model performance metrics.
• MLOps experience (e.g., Docker, Kubernetes, CI/CD for ML, MLflow).
• Cloud experience with Google Cloud Platform (preferred), or AWS/Azure (ML infrastructure, orchestration, storage, and databases).

Nice-to-Haves

• Not explicitly stated, but a research mindset and comfort with ambiguous, high-impact problems are emphasized.