WEX is seeking a highly technical and results-oriented Product Manager to lead the development of our shared AI capabilities. This is a role for a builder. We are looking for a PM who is comfortable opening a code editor to prototype an idea before writing a PRD.
In this role, you will own the lifecycle of autonomous AI agents—from defining the persona and tool capabilities to actively building the proof-of-concepts (POCs) that validate feasibility. You will bridge the gap between business needs and engineering reality, designing AI-powered solutions that serve our diverse lines of business.
1. Technical Discovery & Prototyping (The "Builder" Aspect)
Hands-on Validation: Instead of just defining requirements, you will build functional prototypes (using Python, LangChain, or direct LLM API calls) to validate if a use case is solvable by current models.
Prompt Engineering & Logic: You will own the initial system prompts and logic flows. You understand how to structure inputs to ensure the agent follows instructions and outputs structured data (JSON).
Feasibility Analysis: Translate customer insights into technical hypotheses. You will determine the trade-offs between model intelligence (e.g., GPT-4o) vs. latency and cost (e.g., Llama 3) for specific features.
2. Agent Architecture & Execution
Define Agent Capabilities: specificy the "tools" the AI needs to interact with (e.g., specific internal APIs, SQL databases, or vector stores) to solve the user's problem.
Drive Complex Initiatives: Lead the transition of your prototypes into production-grade software by collaborating with engineering to handle edge cases, rate limits, and security guardrails.
RAG & Context Management: Design how the product retrieves information. You will make decisions on what context needs to be injected into the model to prevent hallucinations.
3. Data, Evals & Performance
LLM Evaluation: Define success beyond just "user happiness." You will implement evaluation frameworks ("Evals") to quantitatively measure agent accuracy, hallucination rates, and successful tool usage.
Insight-Driven Decisions: Use observability tools (like LangSmith or Datadog) to review conversation traces and identify where the AI logic is failing.
4. Collaboration & Influence
Engineer-to-Engineer Communication: You speak the language of the developers. You can conduct code reviews on the logic layer and discuss API specifications for tool calling.
Stakeholder Education: Demystify AI for non-technical stakeholders (Digital, Commercial teams), clearly explaining what the technology can and cannot do to manage expectations effectively.
The "Must-Haves"
Technical Proficiency: Fluency in Python. You must be able to write scripts to test APIs, manipulate data, and interact with LLMs.
Builder Experience: Proven experience building an AI application, chatbot, or agent (side projects count). You have hands-on experience with OpenAI/Anthropic APIs and frameworks like LangChain or LlamaIndex.
Product Experience: Minimum of 2+ years in product management. You know how to prioritize a roadmap, manage a backlog, and focus on business value.
Data Fluency: Ability to write SQL queries to pull your own data and analyze product usage.
The "Nice-to-Haves"
Experience with Vector Databases (Pinecone, Weaviate) and RAG (Retrieval Augmented Generation) architectures.
Experience fine-tuning models for specific tasks.
A Computer Science degree or equivalent software engineering background.