caewerbring@gmail.com | ewerbring.com | Stockholm, Sweden — open to relocation and travel
Summary
Technical founder with 10+ years shipping products in unfamiliar domains — VR games and apps, AR pharma physiotherapy games, real estate operations — repeatedly zero-to-one, from first stakeholder interview to production. Most recently: embedded solo with a property management company and took a manual tenant-placement workflow to a live AI-powered booking system in five weeks, cutting the workload by half (full case study at ewerbring.com). Built an autonomous agent loop that implemented 1,225 commits — 79.3% of the project. MSc Computer Science, KTH Royal Institute of Technology. Slightly on the spectrum, visible professionally in radically direct communication, systematic thinking, and low tolerance for pretending something works when it doesn’t.
Experience
Automation Engineer | Klara Förvaltning (property management)
May — June 2026 | Sweden | Live in production since June 14, 2026
Embedded with a family-owned real estate company to automate their tenant-placement workflow — a manual “booking dance” of email and SMS coordination consuming ~40 hours/month.
- Ran the full engagement solo: discovery interviews (2 days) → workflow specification with flow diagrams validated against employees (1 week) → working prototype with live Twilio/calendar/LLM integrations, role-played with staff (1 week) → autonomous implementation → production
- Shipped an integrated web app that detects contract terminations via the property-management API, matches and contacts applicants, coordinates three-way viewing schedules across employee calendars, and books viewings — with human-in-the-loop oversight and escalation paths preserved throughout
- Hardened the LLM seam: intent routing with structured outputs, deterministic state machines where the LLM underperformed, eval experiments for Swedish multi-actor conversations, and an LLM decision-log tool for production debugging
- Security hardening: PII-redaction scripts, signature-gated Twilio webhooks, baseline vulnerability checks
- Result: system live and booking viewings; ~20 hours/month saved (50% of workload); ~1-year payback on development cost
- Stack: Python/FastAPI, Vue, PostgreSQL + Alembic, Anthropic Claude (conversational intent routing), Twilio SMS, Microsoft Graph calendar, third-party property API, Playwright E2E, Docker, Heroku
- Built “ralph-loops,” a custom autonomous agent development system, and ran up to 7 loops concurrently: the autonomous loop implemented 1,225 commits — 79.3% of the project (2,046 commits in 5 weeks, 99.7% AI-written); test suite grown 63 → 2,196, 92.6% of issues closed within 24h (median 2.4h), peak 75 issues closed in one day
- Published a full technical case study: ewerbring.com/blog/work/five-weeks-to-production
Founder & CEO | Nordic Forest Games
2019 — 2026 | Stockholm, Sweden
Founded exercise game studio. Grew to 8 employees and multi-million-euro total revenue.
- Roche (pharma) — world’s first mobile physiotherapy game: Landed enterprise pharma client at a startup conference. 8-10 month pilot, solo: built computer-vision exercise detection, interviewed doctors and physiotherapists to scope viable exercises for children with spinal muscular atrophy, then traveled across Sweden field-testing with patients — including wheelchair users and children with walking aids — to validate detection in real conditions
- Pilot delivery won a multi-million-euro contract for the full game plus two expansion packs, spanning 2.5-3 years. Client owned the IP; priced accordingly as consultancy
- Planned the engagement in staged gates — technical prototypes → game prototyping → meta-layer → production — so each stage’s output de-risked the next: “designed it so we couldn’t fail”
- Delivered 18/18 milestones on time; Roche said this was a first among their vendors
- Shipped: phone game with meta-game, 5 mini-games each targeting a physiotherapy exercise, analytics reporting for Roche. Tested 15 avatar/game prototypes with kids in playgrounds
- Scaled from solo to 8 employees mid-project; introduced Kanban, standups, retros, 1:1s — first-time manager, learned by reading and doing
- Feel Good Crew (B2C): exercise app with 139,000+ reps logged; power users on daily streaks. Hybrid architecture: Expo/TypeScript main app, embedded Unity module for AR and games, native iOS/Swift integrations (home-screen widget, screen-time limits)
- Patent application: camera-based squat detection (anti-cheat) using front/back camera
- 50+ prototypes across exercise games combining movement detection, computer vision, and game design
AI Development Workflow Evolution
2024 — 2026 | Self-directed
Progressed from ChatGPT copy-paste through Cursor and Windsurf to autonomous Claude Code production pipelines — the journey behind ralph-loops:
- 13-agent GitHub Actions pipeline: PM → Researcher → Architect → TDD Planner/Verifier → Implementer → Validator → Code Reviewer → Visual Tester → Feature Reviewer. Used it to re-architect Feel Good Crew from a pure Unity app into an Expo/TypeScript main app with an embedded Unity game module (AR + games) and native iOS/Swift integrations — moved the app layer out of Unity specifically because agentic development in Unity is subpar: 0 → 25 issues/week over 4 weeks
- Diagnosed its failure mode: worked perfectly with a reference app to port against; became brittle on greenfield features where agents made novel design decisions without a reference
- ralph-loops (current): deterministic bash control flow (AI never routes the workflow), label-based concurrency across parallel loops, planning via parallel sub-agents from different architectural starting points reconciled by a verifier agent into a single checkpointed plan, TDD red-green-refactor execution skill, UI-design-first references so agents self-correct against visual parity, hard timeouts with auto-recovery
- Core insight from iteration: context loss across agent handoffs and individual agent hallucinations are the central failure mode — solved with deterministic engineering at key points and identifying the specific context needed for each task
- Reusable agent infrastructure: maintain a portable template — Docker VM setup, shared credentials, skills, loop scripts — carried from project to project
Co-Founder | Resolution Games
Jan 2015 — Aug 2016 | Stockholm, Sweden
Co-founded what became one of the world’s leading VR game studios. $6M seed round; grew team from 3 to 20.
- Built the company’s first VR prototype on day one; built partner and investor demos throughout
- Produced one of the first games on Google Daydream — managed the team and the Google stakeholder relationship
- Implemented the world’s first virtual reality in-app purchase
- Spoke at SLUSH and Casual Connect on VR
Earlier
- Software Consultant, VR/AR/full-stack (2017-2018)
- CEO, ConsciousCrypto (2018) — retail crypto investment platform
- Founder, Bechno (2016-2017) — VR travel, then VR games
- Co-Founder, Bookl (2013-2014) — pay-per-page ebook service; customer development + full-stack build
- iOS App Developer (2012-2015) — 14 apps, including Syrian language tools, energy calculator, habit app, and more
- Founder, EEA Consulting (2010-2011) — energy cost analysis for industrial and real estate clients
Education
- MSc, Computer Science — KTH Royal Institute of Technology (2012-2015)
- Energy and Computer Science — Mahidol University, Bangkok (2011-2012)
- Energy and Environment Engineering — Linköping University (2009-2011)
- Draper University — Silicon Valley (2013)
- Value Investing — Columbia University (2024)
- Certified Personal Trainer (2025)
- Mandarin — Fudan University, Shanghai (2008-2009)
Technical Skills
AI/Agent Systems: Claude Code & Anthropic API, autonomous agent workflows, multi-agent orchestration, context engineering, prompt engineering, LLM evals, tool-call design, structured outputs, intent routing, human-in-the-loop workflow design, TDD agent loops Development: Python, FastAPI, TypeScript/JavaScript, Vue, React Native/Expo, C#, Unity3D, Swift/Objective-C, PostgreSQL Infrastructure: Docker, Heroku, GitHub Actions, CI/CD, Playwright, Twilio, Microsoft Graph Computer Vision: movement/exercise detection, camera-based input Product: customer discovery, stakeholder interviews, workflow mapping, rapid prototyping, staged-gate project design, Kanban
Recognition
- Forbes 30 Under 30 — Gaming (2016)
- Swedish #1 Innovator — Veckans Affärer (2016)
- Speaker — SLUSH, Casual Connect