Work

Case studies

We work across two practices — production AI engineering, and the brand, visibility, and web work around it. Pick the side relevant to you, or read both.

Applied AI engineering

Production AI systems

RAG, agents, and AI platforms built for regulated industries and on-premises deployment — the systems that have to keep working after the demo.

Enterprise RAG, on-prem

Vermiculus Financial Technology
2025 — 2026

Built an enterprise RAG system from scratch, unifying knowledge across Confluence, Jira, and GitHub. Deployed fully offline on-premises for a regulated finance environment.

  • Hybrid retrieval (BM25 + pgvector) with reciprocal rank fusion
  • Cross-encoder reranking for precision
  • Instructor-XL embeddings and Qwen 32B running locally on NVIDIA GPU
  • REST API, web UI, and Microsoft Teams integration
  • −67% time-to-information across previously siloed sources
RAGOn-prem
Python · pgvector · Qwen 32B · Instructor-XL

AI platform engineering

Walma AI
2026

Engineering on Walma's distributed AI platform — a multi-tenant enterprise system that lets organizations chat against internal knowledge bases and run AI-powered document pipelines. Contributed across four interconnected codebases: Noda (Nuxt chat UI + RAG search and tool execution), Noda Workers (Azure Functions for ingestion and scheduling), Ocle (visual no-code pipeline builder), and Ocle Workers (19-worker serverless processing engine).

  • Invoice Reviewer (Fakturagranskare) — rebuilt a municipal contract auditing workflow from 1 correct / 7 warnings / 3 inconclusive to 11/11 reliable outputs, via a decision-tree prompt with self-check stage, a deterministic extraction pass (temperature 0, fixed seed), retry logic, tool-call caps, and post-processing rules
  • Platform architecture — established and applied the Ocle / Noda separation (stateless reusable document processing vs. customer-specific data, config, and scheduling), connected through a result/handler contract
  • RAG ingestion pipeline — full path from Ocle Workers document parsing → webhook events → embedding generation and Azure AI Search indexing → hybrid search at chat time. Embedding migration from ada-002 to text-embedding-3-small and parent-child chunking for an RCO dataset
  • Infrastructure consolidation — migrated Ocle Workers onto Noda's shared Azure AI Foundry hub, removing per-environment hub provisioning
  • Tenant outcomes include a security-sector knowledge base (−65% support lookup time) and an invoice auditing workflow (−92% manual oversight)
AI platformMulti-tenant
TypeScript · Nuxt 3 · Azure Functions · Azure AI Search · Azure OpenAI · Supabase
Brand, visibility & web

AI visibility, content & commerce

Retrieval-aware content, GEO strategy, analytics, and the web work that makes the rest of it visible — where the audience meets the product.

Content strategy & thought leadership

Walma AI
2026

Built Walma's content presence from the ground up — thought leadership library, LinkedIn strategy, and an analytics-led repositioning. Fully organic, no paid promotion.

  • 29 articles for walma.ai on EU data sovereignty, RAG architecture, shadow AI, the EU AI Act, and production enterprise AI
  • LinkedIn strategy across 14 topic areas in 3 clusters (technical substance · buyer reality · how the team builds), with voice guide, cadence, and a credibility-based measurement framework
  • Diagnosed low organic reach from team analytics and reverse-engineered two viral posts (36K and 124K impressions) into a repeatable structural pattern
  • Over 46 days: +5,766% impressions · +140% page views · +109% unique visitors · +1,050% new followers · search appearances from ~39/week to 800–1,000/week
  • Outperformed a comparable competitor (379 followers) who posted zero times in the same period
ContentStrategy
Content systems · LinkedIn analytics · GEO

Founder advisory · AI visibility & analytics

Independent client
2025 — 2026

LLM visibility strategy and analytics dashboards for a founder-led company. Designed how the brand surfaces inside AI answers and built the dashboards to measure it.

  • Retrieval-aware content architecture and on-page structure
  • Internal dashboards integrating GA, GSC, and Bing Webmaster Tools
  • Measurement framework for AI-driven traffic and brand mentions
GEOStrategy
Next.js · Wagtail · Grafana · Analytics APIs

Direct-to-consumer storefront

Stockholm independent brand
2025

End-to-end direct-to-consumer storefront for an independent Stockholm retailer — brand expression, product catalog, and integration with the operational side of the business.

  • Custom storefront design with clear product narrative
  • Catalog and inventory wired to day-to-day operations
  • Analytics and conversion instrumentation from day one
CommerceWeb
Next.js · Commerce platform · Analytics