This is a remote position.
Role: AI Engineer
Experience: 4–8 years
We are looking for a Senior AI Engineer who treats LLMs as an engineering substrate — someone who
builds production-grade Go services on Google Cloud that turn model output into structured, deterministic,
schema-valid data the rest of the system can trust. This is a hands-on individual-contributor role with
significant ownership over design and implementation.
Our AI/ML work spans several modules — some LLM-backed, some deterministic — and you may contribute
across them over time. We therefore value engineers who are adaptable and strong on fundamentals over
narrow specialists, and who can pick up a new problem space quickly.
Key Responsibilities
• Build & integrate LLMs: Design and build Go services that integrate LLMs into production workflows —
with strict structured output, confidence handling, and deterministic fallbacks when the model is
unavailable or low-confidence.
• Reliable agentic workflows: Build multi-step and agentic workflows that execute reasoning, handle
errors, and maintain state — treating retries, timeouts, rate limiting, and graceful degradation as firstclass concerns.
• Structured & deterministic output: Enforce strict Data systems: Work across the team's data and messaging stack — graph, analytical, and event-driven
stores — modeling data and writing efficient queries.
• Evaluation & reliability: Define and own evaluation for AI components — datasets, regression/eval
harnesses, and metrics for accuracy, latency, cost, and reliability — so prompt and model changes ship
safely.
• Production engineering: Ship multi-tenant, observable services on GCP that meet the team's coding
standard, and review peers' work to the same bar.Mandatory Skills & Qualifications
• Go (Golang), production-grade: Strong, idiomatic Go — concurrency (goroutines, channels, context),
disciplined error handling, and clean, testable service code. You have shipped and maintained Go
backend services in production.
• Applied LLM engineering: Hands-on experience integrating LLMs into production systems — prompt
design, structured/JSON output, function/tool calling, confidence handling, and fallback strategies. A
framework-agnostic grasp of agentic patterns (tool use, multi-step reasoning, state) and why reliability
matters more than cleverness.
• GCP & Vertex AI: Practical experience on Google Cloud, ideally with Vertex AI (Gemini) and common data
and eventing services.
• System-engineering mindset: You approach AI as an engineering problem — idempotency, retries, rate
limiting, timeouts, structured I/O, and graceful degradation rather than just prompt tuning. You design
for observabiData systems: Comfortable with SQL and at least one of: analytical (e.g. BigQuery), graph (e.g. Neo4j /
Cypher), or relational (e.g. PostgreSQL) stores. You can model data and write efficient queries.
• APIs & services: You build clean service interfaces (gRPC / REST / GraphQL) and understand how to
expose backend logic as well-bounded “tools” that AI components can call safely.
Optional (But Highly Valued) Skills
• Python: For prototyping, evaluation tooling, data work, or ML experimentation alongside the primary Go
stack.
• Agent orchestration frameworks: Experience with agent / LLM-orchestration frameworks (e.g. Firebase
Genkit) or comparable tooling.
• Knowledge graphs: Graph modeling, GraphRAG, or relationship inference at scale on graph databases.
• Time-series & ML: Forecasting (e.g. ARIMA and related methods), BigQuery ML, or applied model
evaluation.
• LLM security: Awareness of the OWASP Top 10 for LLM Applications (prompt/query injection, insecure
output handling, excessive agency), particularly where model output drives queries or actions.
• Containerization & delivery: Docker, Kubernetes (GKE), and CI/CD.
Cost & latency optimization: Caching, batching, and model-tier selection to keep AI workloads efficient
at scale.
Tech Stack & Standards
(Experience in these or similar technologies is preferred)
• Language: Go (primary); Python a plus.
• AI / LLM: Vertex AI Gemini; agent-orchestration frameworks (e.g. Firebase Genkit).
• Data & messaging: Graph (e.g. Neo4j / Cypher), analytical (e.g. BigQuery / BigQuery ML), object storage
and event streaming (e.g. Cloud Storage, Pub/Sub).
• Cloud & deployment: Google Cloud Platform; containers and Kubernetes (GKE).
Engineering standards: Multi-tenant isolation, structured error handling, and automated evaluation for
AI components.
• Observability: New Relic or equivalent.
