American Express is looking to build and optimize next-generation AI applications, specifically focusing on model inference, agentic AI systems, embeddings, and clustering to enhance digital banking and payment products.
Requirements
- Proficiency in Python with exposure to AI/ML frameworks (PyTorch, TensorFlow).
- Understanding of embeddings, similarity search, and clustering algorithms.
- Familiarity with inference optimization and deployment concepts (e.g., GPU acceleration, containerization, cloud services).
- Experience with agentic AI frameworks and orchestration tools (LangChain, AutoGen, Haystack).
- Familiarity with vector databases (FAISS, Pinecone, Weaviate) and retrieval-augmented generation (RAG).
- Exposure to MLOps practices (MLflow, Docker, Kubernetes).
- Hands-on experience with LLM APIs (OpenAI, Anthropic, Hugging Face).
Responsibilities
- Assist in building and optimizing inference pipelines for large-scale AI deployments.
- Work with embeddings and clustering methods to enhance retrieval, personalization, and knowledge-structuring use cases.
- Apply techniques such as batching, quantization, and model distillation to improve efficiency.
- Support the design and implementation of agentic AI systems that can reason, plan, and act autonomously.
- Experiment with frameworks (e.g., LangChain, AutoGen) to prototype AI-driven workflows.
- Collaborate on integrating agents with APIs, knowledge bases, and enterprise applications.
- Partner with governance and risk teams to ensure GenAI solutions meet compliance and approval standards.
Other
- Communicate clearly and respectfully with stakeholders across technical and non-technical groups.
- Contribute to internal knowledge-sharing through code reviews, documentation, workshops, and forums.
- Help foster a developer community that encourages experimentation, learning, and responsible innovation.
- Strong communication skills and a collaborative temperament for working with governance and approval teams.
- Enthusiasm for learning and sharing knowledge with peers.
