Call for Engineer

Hermes Web: Call for Founding EngineerHermes Web is looking for a founding engineer to build the first polished, hosted personal AI agent product.Some facts about us:Incubated by Open Core Ventures, the studio founded by Sid Sijbrandij (ex-CEO of GitLab)Building the hosted, non‐technical‐user version of Hermes Agent — the fastest‐growing open‐source agent framework after OpenClawOCV underwrites infrastructure and LLM costs, so you optimize for growth and polish, not unit economicsFully remote, product‐focused, compressed sprint cadenceIf you want to own product and engineering end‐to‐end, ship consumer‐grade polish in a space that's still raw, and help define what a mainstream AI agent product looks like, this is for you.RequirementsHigh agency, ship fast without hand‐holdingStrong Python skills — FastAPI or similar production web framework experience requiredStrong frontend skills (React, Next.js, TypeScript) for the chat, memory, and activity surfacesExperience with auth, billing, and multi‐tenancy in production productsContainer orchestration experience (Kubernetes, Nomad, ECS, or equivalent)Experience with AI‐native tools (Cursor, Claude Code, model APIs)Knowledge of agents, LLM gateways, rate limiting, and abuse preventionComfortable working across frontend, backend, infrastructure, and AI workflowsThe kind of person we want: someone who has shipped a consumer or prosumer product they're still proud of, and who responds to an aggressive timeline with "tight but doable if we cut X, Y, Z" rather than "let me spec it first."Experience with agent frameworks (Hermes, OpenClaw, LangGraph)LLM gateway or proxy workStripe subscription billing at scaleProduction experience hardening agentically‐generated codeOSS contributionsConsumer AI product experience (Cowork, Cursor, or similar)Project: Hosted Hermes Agent Platform1. IntroductionWe are building a multi‐tenant hosted platform that gives anyone their own personal Hermes agent running in the cloud in under 60 seconds. No terminal, no API keys, no configuration. The agent persists, remembers them, and gets better over time.This project includes:Per‐user Hermes containers with network and storage isolationA warm pool provisioning system for instant signup‐to‐first‐messageA production‐hardened chat, memory, and activity interfaceAn LLM gateway that meters usage, enforces caps, and prevents abuseBundled Claude Sonnet access — users never see an API key or model selectorStripe billing infrastructure and self‐serve subscription flows2. RequirementsPlatform and provisioningPer‐tenant container orchestration via Kubernetes or NomadWarm pool pattern for sub‐90‐second provisioningIdle sleep and fast wake (under 2 seconds on incoming message)Persistent per‐tenant memory and conversation storageAuth via Clerk or Auth0 (email + Google OAuth, email verification required)FrontendReact, Next.js, TypeScript, Tailwind, shadcnMulti‐tenant chat, memory, and activity surfacesGraceful handling of connection loss, agent timeouts, container restartsStreaming responses, conversation history, new‐thread flowsLLM gatewayProxy every Anthropic API call from every tenant containerHold Anthropic API keys server‐side; tenant containers never see themPer‐tenant metering with second‐by‐second granularityHard daily and monthly cost capsCircuit breakers for runaway loopsOps kill switch for any tenantBilling and opsStripe subscription integration at $20/moSelf‐serve upgrade, downgrade, cancelInternal ops dashboard for user management and anomaly responseRunbooks for common failure modes3. Non‐GoalsDeveloper‐facing platformSelf‐hosted desktop or mobile appsSkills marketplace or plugin ecosystemWorkflow builder, triggers, scheduled agentsMultiple LLM providers or model selectionBring‐your‐own API keyForking Hermes — we track upstream and contribute back4. Prior ArtHermes Agent — the open‐source framework we build the hosted experience on top ofOpenClaw — the largest agent framework today and reference for community patterns and growth dynamicsLangGraph Deep Agents — architecture patterns for stateful long‐running agent workflowsAnthropic MCP — server design patterns for tool interfaces and permission‐aware contextCowork (Anthropic) — polish bar and UX patterns for non‐technical AI agent productsCursor agent UI — interaction model for AI agents in consumer products#J-18808-Ljbffr