AI Platform Architect

Trading since 2017, Gravity Team is one of the leading crypto market makers and liquidity providers, with cumulative trading volumes to date in excess of $400 billion. We provide 24/7 liquidity across 1,400+ crypto-asset pairs on 30+ exchanges in 15+ countries, representing roughly 1% of global spot trading volume. About the Role We're hiring a senior engineer to design the AI platform that the rest of Gravity will build on — the orchestration, model routing, RAG layer, guardrails, and integrations that turn AI from shadow infrastructure & scattered experiments into a governed enterprise capability. This is a hands-on platform role, not an architecture-diagram role. You’ll join our Head of Business & Artificial Intelligence in a new team to evolve our data foundation into an AI fortress. You'll make the build-vs-buy calls, stand up the stack, operate it in production, and enable non-engineers to ship workflows on top of what you build. You'll own security and governance jointly with our CTO, and partner with the AI adoption lead and business teams to turn high-friction processes into AI-assisted workflows that actually run. Small senior team. Greenfield mandate. Security-first by default. Sky is the limits impact. Responsibilities AI platform strategy — what tools, models, platforms, and patterns we use, when, and why. Build vs. buy vs. adopt decisions, backed by real experience and opinions you can defend. The full stack — evaluate, select, and deploy workflow / orchestration platforms, model providers, vector stores, MCP servers, agent frameworks, observability tooling. End to end. Hands-on build and operations — deployment, scaling, upgrades, integrations with Slack, Atlassian, Databricks, AWS, internal databases. This is engineering, not architecture diagrams. Security and governance (with our vCISO) — secrets management, sandboxing, access controls, data classification, prompt-injection defense, audit logging, human-in-the-loop rules for high-stakes workflows. The knowledge / RAG layer — ingestion pipelines, embedding strategy, retrieval quality, freshness. The foundation everything else sits on. Enablement — reusable templates, primitives, documentation, and debugging support so the AI adoption lead and business users can build workflows on the platform without needing you in the loop. Model access, routing, and cost — API keys, rate limits, per-workflow / team / user cost tracking, model selection per use case, fallback strategies. Platform health — uptime, cost, usage, incidents, security posture. Own the on-call. Required Experience 7+ years in infrastructure / platform engineering / DevOps in production environments. You have shipped and operated internal/external customer-facing systems with high impact. . Hands-on experience building and operating at least one LLM platform in production — workflow orchestration, model routing, RAG, agent frameworks. Not demos. Not POCs. Strong AWS and Docker — VPC, IAM, networking, secrets management, containerized deployments. Kubernetes is a plus, not required. Security mindset by default — secrets handling, least privilege, audit logging, prompt-injection awareness. You've shipped systems where security was non-negotiable. Python and TypeScript fluency — you read source to debug, write integrations, and don't just configure tools. MCP servers, agentic workflows, sanitization, and modern agent orchestration frameworks in production. AI gateway operation — Databricks Unity AI Gateway, Portkey, LiteLLM, AWS Bedrock Guardrails, or equivalent — including rate limiting, audit logging, and policy enforcement. Tracing, evaluation, and lifecycle tooling for LLM / agent systems — MLflow, LangSmith, Weights & Biases, Arize, OpenTelemetry, or equivalent. Must speak to LLM-specific tracing and reproducible eval patterns. Proven track record of enabling non-engineers to ship on platforms you built. Be ready to give specific examples in the interview. Nice to have Regulated-industry background — fintech, crypto, healthcare. You understand audit trails and compliance posture. Crypto / HFT / trading domain knowledge. Spark Clusters, Local/Open-source model experience - Kimi, Llama, Qwen, etc.