Coinbase's Risk AI/ML team needs to manage financial threats impacting the business and users by building scalable and intelligent systems to detect and prevent fraud, scams, and account takeovers at scale, thereby protecting users and enabling future products.
Requirements
- Deep expertise in applied AI/ML techniques (e.g., Risk ML, deep learning, NLP, recommender systems, anomaly detection); with required prior domain experience in Payment Risk, Credit Risk, or Identity Risk / Account Takeover / Scams.
- Expert-level coding skills (e.g., Python) and deep experience with AI/ML frameworks (TensorFlow, PyTorch).
- Experience in building backend systems with a focus on data processing or analytics is a plus.
- Familiarity with modern data and AI/ML infrastructure (e.g., Feature Stores like Tecton, Model Serving like RayServe, Apache Airflow, Spark, Kafka).
- Experience with Graph Neural Networks (GNNs) or Sequential Models (like LSTMs).
- Experience with LLMs (NLP, fine-tuning, agentic systems) or Reinforcement Learning.
- Understanding of MLOps best practices, including monitoring and improving production models.
Responsibilities
- Take full technical ownership of a core problem space, such as Scams or Account Takeover. You will design, build, and lead the strategy for all models in this domain.
- Lead the system design and architecture for new, complex, cross-team risk detection models. This includes everything from feature pipeline design to model selection (e.g., GNNs, LSTMs, LLMs) and high-performance serving.
- You will be the primary technical voice influencing at an organizational level. Work with Product, Ops, and other stakeholders to translate ambiguous business needs into a clear technical roadmap. You will be the primary technical voice defining the "how."
- Act as a senior technical leader and mentor for other senior (IC5), mid-level, and junior engineers on the team.
- Apply modern methodologies (e.g., deep learning, NLP, Graph Neural Networks (GNNs), sequence modeling, and LLMs for NLP and conversational agents) to solve complex, crypto-native challenges.
- Architect the adaptive logic that decides which friction (a quiz, an LLM agent, a human review) to apply to which user, balancing security with user experience.
Other
- In-person participation is required throughout the year.
- Team and company-wide offsites are held multiple times annually to foster collaboration, connection, and alignment. Attendance is expected and fully supported.
- Passion & Values: A commitment to building an open financial system and a strong desire to protect users from fraud and scams. You embody our core cultural values: add positive energy, communicate clearly, be curious, and be builder.
- Team Collaboration: Ability to work collaboratively on technical initiatives and contribute to impactful AI/ML solutions.
- Communication Skills: Strong communication skills, with the ability to convey technical concepts to both technical and non-technical audiences.
