Founding AI Engineer, Lead Applied

About the role

Founding AI engineer: you own the AI platform end to end — architecture, build, ship, measure — and set the technical standards the future AI team will inherit. Applied engineering, not research. Development is run agentically with Claude Code (spec-driven sessions, skills, hooks, MCP tooling) on top of an existing PHP/JS codebase; you should treat AI coding agents as a force multiplier and know exactly when not to trust them.

What you will build

• Conversational agents — messenger-native onboarding and assistance flows (WhatsApp Business API, Telegram, and web chat): deterministic state machines with LLM-powered understanding at the edges, resume-on-return, consent gating, instant-value payoff moments.
• Document intelligence — an extraction platform handling many document types (CVs, certificates, licenses, IDs, employment records) in multiple languages and formats, including photos of paper documents: OCR + LLM extraction into enum-constrained schemas, per-field confidence, golden-corpus testing with accuracy as a CI metric.
• Matching & scoring — LLM-assisted query construction over a deterministic search core, relevance ranking, and profile-quality / reliability scoring models with fairness and explainability built in.
• Verification & trust modules — document validity checks and credential verification workflows (verifiable-credentials standards on the roadmap).
• Agentic & platform infrastructure— the shared tool layer (internal MCP services over platform APIs), evaluation harnesses, guardrails (prompt-injection defense, PII handling, GDPR-clean consent logging), and funnel/loop instrumentation. If it isn’t measured, it didn’t ship.

Requirements

• 5+ years of production software engineering, with strong backend depth (PHP and/or Node.js; part of the job is modernizing around a legacy core).
• Shipped LLM-powered systems in production: tool calling, structured outputs, constrained generation, evaluation pipelines — not prompt experiments.
• Hands-on with agentic development tooling (Claude Code or equivalent) and the judgment to review what it produces.
• Experience with at least two of: conversational/messenger products, document processing / OCR pipelines, search & ranking systems, ML scoring models.
• Strong instincts for data quality and testing: state machines, golden corpora, regression suites for non-deterministic systems.
• Comfortable as the first specialist in a domain: choosing the stack, writing the standards, mentoring as the team grows.
• Working English.

Nice to have

• Multilingual extraction / NLP; low-quality-input robustness (scans, photos).
• Recruiting, HR-tech, or marketplace product experience.
• Growth/analytics engineering (event schemas, funnel dashboards, experimentation).
• Verifiable credentials (W3C VC) or applied cryptography interest.

Apply with a short note about the most interesting system you’ve built and a link to code or a write-up.