Goalscape Backend & AI Infrastructure Lead

What we're building

Goalscape is the intent engineering platform — organisational intent made machine-actionable. Founded 2009 in Kiel, Germany; US-patented visual hierarchy for goal management; seventeen years of compound product wisdom. Today's customers include Volkswagen, TotalEnergies, multiple government agencies, education organisations, and enterprise deployments — alongside thousands of individuals and SMEs worldwide.

With Goalscape Flow — our EU-funded R&D programme — we're building the AI substrate: local LLM topology, MCP wiring, ingestion at production quality, evaluation rigour. Three-year funded runway.

What you'll work on

Goalscape Flow (June 2026 – December 2028)

• Local LLM infrastructure (Ollama; Llama / Mistral / Gemma). Model selection, quantisation, on-prem topology. EU data sovereignty is the product moat.
• Third-party MCP server integration — our local stack talks to external MCPs safely and deliberately.
• Heterogeneous data ingestion — parser framework for PDFs, Excel, ERP exports, CRM data, email, Jira / Git / Asana. Real-world data quality: messy encodings, ambiguous tables, duplicates, partial exports — and validation you can explain to auditors and enterprise buyers, not toy demos.
• A/B-test infrastructure for AI features (cohorts, golden datasets, statistical discipline).
• Cloud-vs-local LLM benchmarking as an explicit deliverable.
• Security audit of the local AI appliance.

Maintained product — backend architecture (not "AI only")

This role includes substantial ownership of the production platform: as lead backend engineer, your scope extends beyond Goalscape Flow's AI work — you will drive hardening and ongoing evolution of the core stack:

• Semi-lattice / synergistic goals — closure-table backing in PostgreSQL; recursive reachability queries; cache coherence (Spring Cache / JCache); Flyway migration hygiene; invariant tests before broader rollout.
• GraphQL — Spring for GraphQL on our core API (schema discipline, DataFetchers, batch loading, subscriptions over WebSockets).
• Real-time collaboration — subscription-driven updates across linked projects; fixing edge cases when goal-link topology changes (product-critical, not polish).
• Per-goal permissions — schema and effective-permission strategy on top of closure tables (design + implementation when lattice stabilises).
• OpenTelemetry across services.
• Modernise our small redirect microservice — Spring Boot 3 / current Java LTS (aligned with parent POM).
• Operational honesty items we'll spell out in the interview: CI consolidation toward GitHub Actions, service-to-service auth clarity, audit logging direction — you'll help prioritise with the lead engineer.

Why this role

Fourth FTE with real architectural imprint. As lead backend, the schema decisions, cache strategy, and GraphQL surface are yours to shape — not someone else's to inherit. Six months in, the next engineer joining will treat your ADRs as canon. AI features and the core platform aren't separate tracks here: one person owns both, because the integration points are where the interesting decisions live.

Who you'll work with

• Marcus Baur, CEO — two-time Olympic sailor, ex-SAP.
• Ivan Klim, Lead Engineering — algorithms, AI integration, codebase depth.
• Richard Parslow — Support & Sales.
• Erik Wehkamp, CGO.
• Advisory board — academics and entrepreneurs.

How we live

Three-year EU-funded runway. Home-office friendly (Kiel preferred for occasional in-person; remote-within-Germany acceptable). Written-first culture (ADRs, design docs). Direct communication.

Compensation

Mid-range starting cash with a documented path to upper-to-top market range tied to growth, plus meaningful VSOPs with unicorn-outcome upside. Full mechanics in interview.

Must-haves

• Senior Java + Spring Boot 3.x (5+ years Java; 3+ years Boot). Spring Security, Spring Data JPA, multi-module Maven.
• PostgreSQL fluency — non-trivial SQL, recursive CTEs, query plans, index design. Closure-table / DAG / reachability patterns a strong plus.
• Server-side GraphQL — Spring for GraphQL: schema wiring, DataFetchers/batch loading, subscriptions.
• Long-running backend architecture — you've shipped systems where correctness under concurrency, cache invalidation, and migration safety mattered; you can point to one concrete example.
• Local LLM operations in production (Ollama, vLLM, llama.cpp, or equivalent). Open-source model families (Llama, Mistral, Gemma).
• MCP server implementation experience (or demonstrably adjacent — clear in your application).
• LLM features in production — observability, cost telemetry, prompt versioning, fallback behaviour.
• Heterogeneous data parsing / ingestion frameworks — you've shipped pipelines beyond happy-path CSV (PDFs, spreadsheets, enterprise exports, email or ticket dumps — pick what matches your CV).
• A/B-test infrastructure for AI features (or strong statistical/experiment design adjacent).
• German residency (hard constraint — EU/EFRE grant rule).

Strong pluses (not all required)

• OAuth2 resource server + Keycloak-adjacent integration depth (custom plugins / admin API awareness).
• WebFlux / reactive HTTP clients for streaming AI calls (cancellation, back-pressure).
• Flyway discipline at scale; performance benchmarking of graph-shaped workloads.
• OpenTelemetry end-to-end (trace propagation across HTTP + WebSocket boundaries).

Cultural fit

You treat AI tools as a multiplier while staying in the driver's seat. You've shipped LLM paths with cost/latency/quality trade-offs — not demos alone. You're comfortable with behaviour preservation for long-tenure enterprise customers. You write things down. You're honest about reservations — we'll be honest about scope.