10a Labs is a safety and threat-intelligence layer trusted by leading AI labs and technology platforms. They are seeking an experienced Machine Learning Engineer to own the full ML lifecycle, from data pipelines to model deployment, ensuring the robustness and scalability of real-world ML systems.
Responsibilities:
- Build and deploy a multi-stage classification system optimized for high throughput and low latency, while ensuring high recall and precision
- Integrate continuous feedback loops from human review to refine model performance
- Design and implement real-world ML systems with a focus on robustness, observability, and scalability
- Collaborate with researchers and SMEs to generate training data and test against edge cases
- Work closely with a broader team of engineers to integrate ML components into production systems and ensure end-to-end system performance
Requirements:
- At least 3–8+ years of professional working experience as a Machine Learning engineer, building, owning and deploying machine learning systems in production
- Strong foundation in traditional ML techniques (e.g., clustering, anomaly detection, supervised learning)
- Hands-on experience with LLMs (e.g., OpenAI, Claude, LLaMA), including fine-tuning and prompt engineering
- Proficiency in Python and modern ML / NLP tooling
- Experience training models on small datasets and using in-context learning techniques
- Familiarity with text processing pipelines, semantic embeddings, and vector search
- Clear communicator of complex technical concepts to non-technical audiences
- Experience deploying models in cloud environments (e.g., AWS, GCP)
- Experience designing or integrating human-in-the-loop systems for model evaluation or policy alignment
- Real-time ML pipelines
- Scaled moderation or large-scale threat detection
- Vision, audio, OCR, or deepfake classification
- Designing multilingual embedding systems with code-switch detection
- Agentic pipelines for explainable or rationale-based moderation
- Rapid prototyping using modern LLM APIs and frameworks (e.g., OpenAI, Hugging Face, LangChain)
- Error analysis and model forensics—comfortable diving into false positives and failure modes