Forward Deployed AI Engineer

Tenarai - LATAM; São Paulo; On-site; BRL 320,000 - 420,000; Full time

Job summary

Tenarai places small, senior engineering teams inside large enterprises and builds alongside the client's own team. This Forward Deployed AI Engineer role runs a governance-focused knowledge graph and AI-agent data platform for a major fintech client.

You will own a use case end to end, working with client business teams to define metrics, entities, and questions, and then build, test, and deploy the retrieval and agent chain.

Qualifications

• Several years of software engineering with strong problem-solving skills and Python fluency.
• Hands-on experience building LLM or agent applications, including retrieval-augmented generation and tool or function calling.

Responsibilities

• Capture client-defined metrics, entities, and questions and convert them into structured, governed context.
• Build the knowledge graph slice and retrieval behind an AI agent.
• Stand up evaluation checks and drive accuracy to target before deployment.
• Reduce time from a new question to a trusted answer; typically weeks/months to days.
• Hand over the work to the client’s stewards for ongoing maintenance.
• Share learnings with Tenarai's platform and graph teams to improve future use cases.

Skills

Python

LLM applications

RAG

APIs

Job description

About the role

Tenarai places small, senior engineering teams inside large enterprises and builds alongside the client's own people. This role is for a current engagement: a program at a major financial-technology company to build a governed knowledge graph and an AI-agent data platform — a single, trusted source of business meaning that the company's AI agents and analytics read from.As a Forward Deployed AI Engineer you own a use case end to end. You sit with a client business team, learn their domain, turn their definitions and data into governed context, build the graph and retrieval behind it, put it behind an agent, and prove the answers are reliable. You are the bridge between the client's domain experts and Tenarai's graph and platform specialists.

What you'll do

• Work directly with a client business team to capture how they define their metrics, entities, and questions, and turn that into structured, governed context.
• Build the slice of the knowledge graph and retrieval that a use case needs, and put it behind an AI agent that answers real questions.
• Stand up evaluation checks for your use case and drive the answer accuracy up to the agreed target before anything ships.
• Cut the time from a new question to a reliable answer, which in these programs typically starts at weeks or months and is expected to reach days.
• Hand the work over so the client's own stewards can maintain and extend it after you move to the next use case.
• Feed what you learn back to Tenarai's platform and graph teams so the shared foundation improves with each use case.

Your first 90 days

• In the first month, pick one use case the client already has underway, learn its domain, and model a single business entity end to end.
• By day 60, expose that entity through a retrieval interface and an agent, and run it against real questions rather than a demo.
• By day 90, have evaluation gates in place, meet the accuracy target, and show a measurable drop in time to a trusted answer.

What you'll bring

• Several years of software engineering with strong general problem-solving and fluency in Python.
• Hands-on experience building LLM or agent applications, including retrieval-augmented generation, tool or function calling, and agent tool protocols such as MCP.
• Comfort embedding with a client team: listening to domain experts, explaining trade-offs plainly, and shipping in someone else's environment.
• A working grasp of data modelling and APIs, enough to design the shape of the data a use case needs.
• The habit of measuring your own output — writing evaluations and acting on what they show.

Helpful, not required

• Experience with knowledge graphs or semantic data models.
• Work in a financial, accounting, or other regulated domain.
• Experience taking a feature from prototype to production with real reliability targets.
• Prior consulting or forward-deployed work with external clients.

How we'll measure success

In the first two quarters, a use case you own is live and meeting its agreed accuracy target, the time from a new question to a trusted answer has dropped measurably, and a client steward can maintain the work without you.

How we work

You will spend most of your week with the client team, with the rest of Tenarai's graph, architecture, and evaluation specialists a message away. We keep teams small, review each other's work, and prefer shipping a thin slice end to end over building in isolation. Several years of software engineering with strong general problem-solving and fluency in Python., Hands-on experience building LLM or agent applications, including retrieval-augmented generation, tool or function calling, and agent tool protocols such as MCP., Comfort embedding with a client team: listening to domain experts, explaining trade-offs plainly, and shipping in someone else's environment., A working grasp of data modelling and APIs, enough to design the shape of the data a use case needs., The habit of measuring your own output — writing evaluations and acting on what they show.