Frequently Asked Questions
Short answers about my work and where to read more on this site. For depth, exploreprojectsandwriting.
Engineer and founder building across the stack: data platforms, AI automation and agents — and the business conversations that decide what's worth building.
Questions below are grouped by data platforms, AI automation & agents, product & business, and a few interests beyond day job work.
About this site
Who is Emil Ingemar Karlsson?
What is this site?
Where should I start?
The AI timeline on the home page, then one case per area: Enterprise data platform, The Unnamed Roads, and B2B platform across 35 markets.
Data platforms
What kind of data platforms have you built?
Governed lakehouse platforms on Azure Databricks (medallion architecture, Unity Catalog, production ML and NLP), plus earlier operations analytics and BI work. Flagship case: enterprise ML and data platform.
Which data tools do you use?
Databricks, Snowflake, DuckDB/MotherDuck, SQL, Python, Power BI, and MLflow in production settings. See the toolstack for the full list.
AI automation & agents
What is agentic engineering, as you use the term?
Designing, building, and operating multi-agent systems that run in production — goal-directed autonomy, handoffs between agents, and human gates where risk matters. Not the same as using a coding assistant. Full definition: what is agentic engineering. Layers: Company OS vs framework vs orchestration.
How do you keep agent costs and risk under control?
Route each task class to the cheapest capable model and the right agent role; promote or demote with evals when quality drifts; use approvals and spend policies before anything ships or spends. Start at the cost/ROI series hub and Operating Note 002.
Automation or agents: when do you use which?
Deterministic workflows (n8n, event-driven functions) when steps are known and repeatable. Agents when judgment, tool selection, or multi-step reasoning is required — always with scoped tools and human oversight where it matters.
What is Company OS?
My working thesis for coordinating intent, context, agents, tools, approvals, and outcomes — not a product or industry standard. Architecture and vision: Company OS architecture vision plan. What changed in practice: Company OS learnings.
What evidence backs your multi-project agent org and infra band?
I operate 10+ live projects on 30+ containers with 12+ MCP integrations — infrastructure around ~€35/month (LLM APIs billed separately). Case narrative: autonomous content org on €35/month.
Product & business
How do product and business decisions shape what you build?
Platform investment and product direction need a shared evidence base — not only stakeholder opinion. Cases: B2B platform across 35 markets, founding TUVA, and failure-priced ROI as a design constraint for agents.
Beyond work
What's the hockey analytics about?
SHL and NHL data pipelines, analysis with hockey context, and publishing at thehockeyanalytics.com. Start with the Hockey Analytics case study, then LINHAC 2026 and SHL cores vs coaching changes.
What's in the Library?
Books, podcast episodes, and YouTube channels I follow — a less formal view of ideas that shape my thinking. Browse the Library.
For questions about anything on the site: contact.