10a Labs is looking to apply machine learning techniques to high-impact research problems, specifically in the areas of AI security, abuse detection, and red teaming of AI systems. The role aims to develop and refine ML models for these critical challenges.
Requirements
- Experience with NLP foundations and text data processing, including cleaning, tokenization, and feature engineering for downstream model development.
- Strong Python background with practical experience using multiple ML frameworks (PyTorch, TensorFlow, scikit-learn, etc.) to prototype, train, and evaluate models in real-world applications.
- Practical experience in generative model adaptation through fine-tuning, prompt-engineering, and in-context learning on low-resource or specialized datasets
- Strong understanding of modeling concepts including bias-variance tradeoffs, regularization, generalization, data imbalance, and model calibration to design models in challenging problem spaces.
- Computer vision skills (OCR, image classification, deep fake detection).
- Experience working in cloud environments such as AWS or GCP for end-to-end ML workflows, including model training, deployment, and monitoring using tools like Vertex AI, SageMaker, and cloud-native ML libraries.
- Understanding of modern retrieval-augmented generation (RAG), AI agent frameworks, and context-aware orchestration (e.g., LangChain, LlamaIndex, OpenAI Agents, or AutoGen) for building intelligent applications.
Responsibilities
- Assist in the development and maintenance of bespoke classification systems, contributing across the project lifecycle.
- Collaborate with engineers across disciplines to co-design project timelines for model development, gaining hands-on experience in planning data pipelines, building infrastructure, deploying models, and assessing performance/latency metrics.
- Research new approaches to help streamline existing processes with the help of machine learning algorithms.
- Assist with research experiment design and automation, particularly as it relates to abuse detection or red teaming of AI systems.
- Ideate / brainstorm new research approaches to known and novel problems in the Trust & Safety and AI Security fields.
Other
- Brings curiosity and creativity to ambiguous research problems, with a bias toward experimentation and rapid iteration.
- Thrives in collaborative, interdisciplinary environments; is resourceful, proactive, and adaptable.
- Is comfortable communicating technical ideas clearly to both technical and non-technical audiences.
- Is excited about contributing to real-world applications of ML and exploring new methods that push beyond standard benchmarks.
- Remote work (based in the continentalU.S.)
