Role overview
We are looking to expand our team with two additional Generative AI Engineers to help build the next generation of psychological assessment systems. The systems we are building uses a graph-based orchestration architecture with a governed state store, retrieval-grounded intervention delivery, and a digital twin layer that models individual psychological profiles through non-intrusive measures over time. We are at an advanced stage of development. You will not be starting from scratch.
We have two positions open. Both are engineering roles with significant depth requirements. One position has a stronger emphasis on the assessment and intervention delivery layer. The other is focused on the digital twinning architecture. You will work closely with AI systems psychologists. Day-to-day tasks will include designing, implementing and evaluating multi-agent architectures, assisting with fine-tuning machine learning models, collaborating with multidisciplinary experts, and utilizing cutting-edge AI techniques to create impactful, science-based psychological assessment and development tools.
What you'll work on
- Build graph or state-machine orchestrator agents that enforces a structured multi-step workflow with persistence and resumability for long-running interactions and handoffs.
- Implement RAG pipeline for a curated content and intervention library, including ingestion, chunking, metadata design, hybrid retrieval, reranking, and provenance.
- Design scoped context layer that exposes only policy-approved, banded user state to the model while keeping the user experience seamless.
- Implement safety controls and escalation flows aligned to modern LLM threat models, including prompt injection and sensitive data leakage risks.
- Instrument the system for evaluation, monitoring, and regression testing so changes do not degrade safety or retrieval quality.
- Fine-tune Llama-based models for psychological state, and trait classifications
- Additional engineering tasks appropriate to the nature and scope of work from our clients
What we're looking for
Applications must include a portfolio with at least two of the following:
- An orchestrated agent workflow (graph or state machine)
- A production RAG system (retrieval design, metadata, evaluation)
- A concrete example of an LLM safety mitigation you implemented (threat model, controls, testing)
- Describe a graph-based orchestration you implemented and how you handled retries, persistence, and evaluation
Interview process
The selection process has three stages:
- Portfolio and CV review. We will assess your portfolio and application materials before scheduling any conversations.
- Technical interview. A structured conversation with our Chief Solutions Architect. Expect questions on architecture decisions, failure modes, and how you think about safety in psychologically sensitive contexts.
- Technical project and assessments. Shortlisted candidates will complete a time-limited technical exercise and a set of assessments relevant to the role.