Daniel Nyström

IAM-guarded MCP server giving AI assistants read access to training, sleep, HRV and recovery data from Garmin Connect, plus goal setting and weekly reviews. Sensitive personal data is only reachable behind authorization; credentials are entered solely through a separate, access-controlled web page.

• Designed for data minimization: read-only against Garmin and no stored health history in v1

Bookkeeping SaaS for Swedish small businesses with an AI assistant that reads, analyzes and books through typed tools. Rust backend (Axum, async-graphql, SQLx); authentication and authorization (domain-scoped RBAC) are split into a dedicated Rust IAM service for security. Live (invite only).

• Built an LLM agent and an MCP server in Rust with typed tools for the general ledger, journal entries, bank reconciliation, balance sheet and income statement, per diem and driving logs
• Production guardrails: tools scoped per company and user, opaque references instead of database IDs exposed to the model, rate limiting and token tracking – and write actions (e.g. booking mileage allowance) require user confirmation
• Built IAM as an isolated authentication service in Rust (domain-scoped RBAC, OAuth, SDK packages for TypeScript and Rust) – auth separated from the application for security
• Testable AI integration: mocked LLM client for deterministic tests and integration tests for the MCP tools

Rust GraphQL API and React SSR frontend for dive sites, with S3-based photo storage. The MCP server lets AI assistants search dive sites, calculate gas blends and maintain site data – write access requires domain permissions.

Hosting platform for all my own products: a cloud k3s cluster with GitOps via ArgoCD and Kustomize, cert-manager, ingress-nginx, sealed secrets and self-hosted PostgreSQL, Redis, MinIO, MQTT, container registry and observability (GlitchTip, Umami). Local mirror in minikube and self-hosted CI runners.

Worked in a team responsible for the Apollo Federation Router and car configuration API in an enterprise-scale system.

• Cut response times and ensured real-time consistency in an enterprise-scale car configuration system through Kafka-based cache invalidation
• Drove architecture improvements in the Apollo Federation router through close collaboration across teams

Delivered IoT solutions for real-time positioning and localization in heavy industry.

• Enabled low-latency real-time positioning in heavy industry through high-performance Rust-based backend services over MQTT
• Built a web-based configuration interface in React/TypeScript with GraphQL for system administration

Built an economy engine for games (proof of concept) in Rust and TypeScript, with domain modeling of the game economy and blockchain integration on Solana.

• Laid the technical foundation with a monorepo architecture and backend/frontend for the game economy
• Translated customer requirements into technical solutions, ran code reviews and led agile ceremonies
• Implemented token and NFT minting on Solana and wrote smart contracts in Rust

Led development of a proof-of-concept for collecting and enriching data from external services.

• Technical lead for a data enrichment PoC with competitive benchmarking
• Investigated and designed architecture for an event-sourcing migration from a legacy system

Java developer in an agile team building a calculation engine with a microservices architecture.

• Built a calculation engine with microservices and the SAGA pattern for reliable distributed transactions

Modernized a legacy web application in a small team.

• Migrated a legacy web app from Kendo to React for improved performance and maintainability