Founding AI/ML Engineer

Xuman.AI; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Benefits offered by this job

Equity stake

Early-stage cash compensation

Milestone-based growth

Job summary

Xuman.AI is building the Human Experience Engine and is hiring for a role governing the agentic layer of our live product. You will design, run, and rebuild the multi-agent system, including tool calling, orchestration, guardrails, and voice on LiveKit, with real users interacting with each agent.

You will own model strategy and costs, and shape the data/ML foundation, collaborating with two AI engineers to ship production-ready AI experiences at scale in a rapidly evolving startup environment.

Qualifications

• 5+ years of real AI/ML work with agentic LLM products shipped to users.
• A model you trained or fine-tuned and production-ready with hand-built evals.
• Multi-agent orchestration with failure paths and architecture changes.
• Python in production and comfort with a TypeScript service (NestJS gateway).
• Daily use of AI tools (Claude, Copilot, ChatGPT) to write, debug, and ship faster.

Responsibilities

• Own the agentic layer in production: agents, tool calling, orchestration, guardrails, voice on LiveKit.
• Define model strategy and model cost: what we run, where, at what price.
• Own the data and ML foundation, collaborating with two AI engineers.

Skills

AI/ML experience

Agentic LLM products shipped

Python production

TypeScript

Multi-agent orchestration

Model evaluation

Tools

NestJS

Job description

Early-stage cash, meaningful equity, and growth tied to the milestones we hit.

Live on both stores. Built to connect billions.

Xuman.AI is a social app where people post content about what they are open to sharing and experiencing. Someone can discover them by scrolling, and if it resonates and vibes with them, they can connect instantly.

On the surface, it collapses fragmented workflows into one system.

Underneath, we are building something far deeper: The Human Experience Engine, our Persona Compatibility layer. This engine matches the right person with the right fit at the right time, based on who they are, what they want, and how they connect.

The founder's vision is to layer cognitive science into human persona compatibility, leading to hyper-personalized, real-time human connections.

Xuman.AI is the productization of the founder's previous bootstrapped venture: ideation, market research, pilots, MVP, and a technical foundation built to scale. The product is built and live on both app stores. We are past the building phase and into distribution: getting it in front of real users and turning it into a repeatable business.

Why this role matters:

The agentic layer is built: agents, tool calling, orchestration, guardrails, voice on LiveKit. All of it is yours to run in front of real users. Models move monthly and you decide what we run.

What you'll own:

• The agentic layer in production: agents, tool calling, orchestration, guardrails, voice on LiveKit
• Model strategy and model cost: what we run, where, at what price
• The data and ML foundation, and two AI engineers

Must have:

• Five years or more of real AI or ML work, agentic LLM products shipped to users
• A model you trained or fine-tuned and put in production, with evals you built by hand
• Multi-agent orchestration with failure paths, and an architecture you changed after last quarter shipped
• Python in production and comfortable in a TypeScript service; ours is a NestJS gateway
• You use AI tools (Claude, Copilot, ChatGPT) daily to write, debug, and ship faster. Not optional; this is how we work.

What you'll get:

• The agentic layer of a live product, yours to design, run and rebuild
• A founding stake: meaningful equity alongside early-stage cash, and growth tied to the milestones we hit
• Real users on the other side of every agent you ship

You are joining before growth, not after. What you own here is real, and it grows as fast as you do.