Private Draft

The 29 personas behind AI

We’ve organized every stage and persona in the AI supply chain, informed by real recruiting at frontier companies. Click any row to see matching profiles from our talent graph.

Shaped by Industry Experts

← ATOMS & ENERGYUSERS & MARKETS →

← Back

Full-Stack Builder

Ships products with models

Full-Stack Builder

Summary

Known as: Software Engineer, AI Engineer, LLM Engineer, Agent Engineer, Product Engineer

End-to-end builder who ships AI products by assembling existing capabilities instead of training from scratch. Cross-functional and multidisciplinary, spanning product, UX, and engineering to take ideas from concept to production (rarely authoring code by hand). Practices AI engineering: context management, light fine-tuning, and model selection (the taste that blends cost, capability, and latency). Shapes AI behavior without touching weights: system prompts, tool-use policies, multi-model orchestration, agentic guardrails, and tight evaluation loops.

Specializations

Agent Engineering — Builds and ships agentic software end-to-end: tool routing and design, prompt engineering, state and context management, memory interfaces, planning and execution loops, and domain workflows. Makes agents robust in production: failure handling and graceful degradation in agent workflows, eval-driven hardening against long-tail failures, and iterative quality tightening. Owns outcome-based evals, LLM-as-judge grading, and trace-level debugging to close the loop between agent behavior and quality. For multi-agent systems, includes orchestration patterns (delegation, subagent contracts, coordination heuristics). Tool-description quality, context window management, and lightweight eval-iteration tooling are the highest-leverage work.

Retrieval & Integration — RAG pipelines, embeddings, vector search, data connectors, and context engineering that ground model outputs in real data. Manages chunking strategies, retrieval ranking, permissions-aware filtering, and the latency/quality tradeoffs that determine whether retrieved context actually helps. At product companies, often the largest share of day-to-day builder work — wiring knowledge bases, APIs, and enterprise data into AI features that return accurate, sourced answers.

AI Interaction Design (HCI) — Technical design of human-in-the-loop systems: interaction models (turn-taking, grounding, clarification), UI affordances for uncertainty and provenance, error recovery and safe fallbacks, latency and interruption handling, and instrumentation for learning loops (telemetry, experiments, rubric-based evals). Often prototypes quickly in code, couples prompt and tool design with UI, and uses lightweight user research methods to validate behaviors under real workflows. Commonly shows up as Design Engineer, AI UX Engineer, or Conversational UX (technical).

Where the Work Lives

[1]Substrate

[2]Compute

[3]Intelligence

[4]Systems

Primary

Assembles models, tools, and infrastructure into reliable production AI features and agentic workflows.

[5]Distribution

Primary

Owns the product surfaces where AI capabilities reach end users.

Candidate Archetypes