Head of AI Engineering
About levelheaded
Levelheaded is reimagining how disputes get solved at scale. We blend humans and AI to power a resolution process that’s fast, fair, and human. Our AI engine, lev, is not a feature—it’s the connective tissue of our platform: conversations, suggesting possible paths to resolution, and helping people move on with life.
We're not building legal tech. We’re building resolution infrastructure and a brand new category of solving problems at scale with a native AI / LLM build.
Why This Role Matters
We're looking for a hands-on, high-context Head of AI Engineering to own the architecture, framework and delivery of our multi-agent AI system. This is a high-performing player-coach role for someone who thrives in ambiguity, builds quickly, and loves turning human problems into structured, scalable AI native systems.
For the right candidate, this role has a clear track to CTO.
What You’ll Do
Own the AI system architecture and build. Including multi-agent workflows for our intake, tone scoring, sync and async resolution paths, and outcome generation/training
Build and deploy custom agents using Claude, RAG pipelines, and CrewAI, among others, specific to our current ICPs
Write Python and orchestration code for LLMs, while supporting backend integrations (Node.js, Postgres, AWS)
Lead prompt versioning, real-time evaluation, and HITL escalation logic
Partner deeply with product to translate real problems into useful, empathetic AI workflows that scale fast, have training and feedback loops and create unique competitive advantage in today’s fast-moving legal tech space
Extend and optimize our RAG pipelines to be emotion-aware and issue-specific
Build internal tooling to monitor prompt health, system trust, and conversation-level insights
Collaborate with devs, PM, UX, and founders to ship product with speed and precision
Define and grow the engineering culture that will carry this platform to scale
You Might Be a Fit If…
You have 8–12+ years engineering experience, with 3+ years leading AI/LLM-heavy builds
You’ve built and shipped AI-first products (ideally using CrewAI, LangGraph, or RAG stacks)
You’re fluent in Python, Claude/OpenAI APIs, vector DBs, prompt engineering, and eval tooling
You’ve worked full-stack or know how to partner with frontend/backend teams
You’re opinionated about HITL, agentic orchestration, and vertical model fine-tuning
You care deeply about outcomes, not just building models
You’re excited to help architect our moat using proprietary tone, resolution, and feedback data in a category we are defining as the business scales
You can mentor other devs while staying hands-on, curious, and fast-moving
Success Looks Like: First 3–6 Months
Clear plan & strategy for our architecture to create a moat, feedback loops, fine-tuning, and proprietary data capture
Three core ‘lev’ agents shipped (Intake, Session/Mediation, Resolution) live and assisting in production cases at scale, fine-tune / train our ‘universal’ lev intake agent, our initial handshake agent
75%+ of low-complexity disputes handled ‘AI-first’, with HITL in place where most impactful
Live async resolution preview flow integrated into our product
Tone monitor, trust meter, and prompt health tooling deployed
RAG pipelines optimized for the top 5 property dispute types
What We’re Building (and Why It Matters)
From Mo, our CEO:
"We’re not building a legaltech platform or feature … we’re creating a brand new category and a resolution engine that solves problems at scale.
We believe empathy and AI can coexist. Lev isn’t just AI automation; it’s how we scale access to justice.”
Current Stack
AI/LLM: Anthropic Claude, OpenAI, Tool Usage, Agents, Prompt engineering
Backend: Node.js, Postgres, AWS (Lambda, S3, SES, Bedrock)
Frontend: Next.js, React, TailwindCSS
Infra: GitHub Actions for CI/CD
Bonus If You’ve…
Built agentic workflows across legal or similar verticals
Led and or built a team of AI engineers in early-stage startup environments
Experimented with RLHF or agent-based simulation
Been part of a platform company that scaled fast without losing soul and created a defensible, deep data-driven moat to secure market leadership and future scale