Resume

I am a lead engineer who builds systems end to end, from core architecture to cloud and applied AI. I favor simple, maintainable solutions over clever code. I explain technical trade-offs plainly to stakeholders and enjoy growing other developers through regular pairing and honest feedback.

Experience

• Nov 2025 - Present
  Lead AI & Software Architect | Norclin (Vetdux)

  • Led architecture and delivery of Vetdux, an AI-driven anamnesis platform for veterinary care, end to end: product, LLM engine, backend, two web apps, cloud, security and compliance.
  • Built a 14-node LangGraph interview engine with parallel and speculative LLM execution; with the later optimisations, worst-case per-turn latency fell from 20-60 s to about 1-5 s.
  • Designed an LLM evaluation and load-test harness: 2,689 synthetic clinical cases, 106 adversarial across 13 attack types, 1000+ sessions.
  • Documented the system to an unusual standard: 23,000+ lines of design docs, 15+ runbooks, and decision records with rejected alternatives.
  • Built multi-account AWS plus GCP Vertex AI entirely as code with Terraform: pausable stacks and 13+ reusable modules.
  • Read the full case study.

• Aug 2022 - Present
  Co-Founder & Lead Engineer | Parabel Game Studio

  • Co-founded the studio and led Aeriasong, a 2D action roguelike for PC, from an empty project to a release-ready build on Steam.
  • Set the technical direction for four years: adopted Godot 4 in alpha, decided where C++ earned its cost, and designed the core architecture.
  • Mentored a second developer from the first week through pair programming and code review, then handed them ownership of whole subsystems.
  • Delivered 100+ abilities, 21 enemy types, three boss fights and three procedurally generated worlds with a two-person team.

• Apr 2015 - Aug 2015
  Engineering Consultant | Midroc (on assignment at Forsmark)

• Sep 2014 - Dec 2014
  Engineering Intern | Forsmark Kraftgrupp AB

Education

2016 - 2022 Uppsala University

Engineering Physics (Teknisk Fysik), not completed

Projects

• Vetdux - LLM Evaluation & Load-Test Harness

  • LLM eval and load-test harness: 2,689 synthetic clinical cases, 106 adversarial attacks.
  • Corpus generated by multi-agent LLM workflows with hard coverage gates that fail loudly.
  • Load generation with token-bucket rate control, p50/p95 reporting, 1000+ session target.

  AI/LLMs, LLM Evaluation, JavaScript, GraphQL, Gemini, Load Testing

• Vetdux - GDPR & EU AI Act Compliance Framework

  • GDPR and EU AI Act compliance framework for an AI platform, implemented as working code.
  • Seven-tier data retention with a legal basis per tier, enforced by background workers.
  • Anonymization gated on k-anonymity, l-diversity and an LLM motivated-intruder probe.

  GDPR, EU AI Act, Privacy Engineering, Python, PostgreSQL, AWS

• Wedding Website - saraoliver.se

  • Designed, built, and launched a full-featured wedding website with RSVP and accommodation booking, including per-room capacity limits and automatic per-person cost splitting.
  • Implemented two-tier access control: a site-wide invite gate with HMAC-signed cookies and constant-time verification, plus a Discord-OAuth admin portal for managing submissions.
  • GDPR by design: Article 9 consent for dietary data and no cookie banner needed.

  Next.js, TypeScript, React, Prisma, Turso, Better Auth, Zod, Tailwind CSS, Vercel, GDPR

• Vetdux - Product Scoping & Business Case

  • Scoped an AI veterinary product into seven priced modules and wrote the commercial offer.
  • Build-versus-buy analysis with explicit cost comparisons for every managed service.

  Product Strategy, Business Analysis, Requirements Engineering, Technical Writing

Skills