Founding Software Architect (Mission Control) - Fulltime - US-based

You are the first engineering hire. You own the technical design of the Mission Control core and make the final engineering calls. You build hands-on from day one, and you lead a team of about eight engineers (backend / platform, AI / agent, infrastructure / security, full-stack) as they join.

The target: a working core on all three deployment targets within the first three months, then three production templates and first design-partner missions by month six.

What you’ll do

• Turn our architecture requirements and prototype PRD into the system architecture
• Make the open stack decisions: agent framework (e.g. LangGraph, NVIDIA AI-Q / NeMo Agent Toolkit or alternatives), durable execution (e.g. Temporal), multi-tenant sandboxing, evidence-ledger storage and GPU sizing for open-weight models
• Design and build the core: the mission object and lifecycle state machine, approval gates (G0-G6) enforced as policy-as-data, a kill switch, the planner and configurable agent roles, checkpoint / resume for missions that run hours to weeks, and budget metering
• Design the append-only evidence ledger, claim-to-evidence tracing and an independent verification layer that enforces our evidence standard
• Architect one codebase for three targets: multi-tenant hosted with per-tenant keys and strict isolation; single-tenant client cloud with client-held keys; air-gapped with offline install and update packages and zero outbound network calls
• Build the model gateway that routes work across frontier and open-weight models by capability, cost, data class and policy
• Set coding, security and review standards; design toward SOC 2 / ISO 27001 from day one, and keep the architecture ready for export-control (ITAR) requirements: US-persons-only access, US-resident compute and FIPS-validated encryption as configurable policy
• Hire and lead the engineering team through the core build, working closely with the product / operations lead and our forward-deployed engineers

• Staff-level engineering experience (typically 8+ years), including technical leadership of a team
• Has shipped and operated production distributed systems
• Has led the architecture of a multi-tenant SaaS product, including tenant isolation and per-tenant data security
• Hands-on experience building LLM agent systems in production: orchestration, tool use, evaluation
• Strong Kubernetes / container experience; designs for portability across cloud providers
• Solid security fundamentals: RBAC / SSO, encryption and key management, audit logging, tenant isolation
• Wants to build hands-on at a pre-seed company, not only direct others
• Based in the US, * Has architected an on-prem, air-gapped or single-tenant self-hosted product (offline install and updates, customer-held keys)
• Experience at a forward-deployed-engineering (FDE) model company
• Deployments in regulated industries: government, defense, critical infrastructure (FedRAMP, ITAR, SOC 2)
• Durable workflow engines (e.g. Temporal) and append-only or event-sourced data stores
• Serving open-weight models on customer GPUs
• Domain exposure to robotics, data centers / construction or aerospace manufacturing
• Eligible to access export-controlled (ITAR) technical data

Mission Control is the AI backend of a human-led, service-as-software company. A client submits a mission (a goal, their data, rules and constraints). A governed team of AI agents then plans, researches, analyzes, simulates, challenges and assembles a verified, evidence-backed decision package. Every material claim traces back to evidence, and an expert partner signs off before anything is delivered.