Quality over quantity.

Our aim is to hire only the top ten per cent in the industry — and the same bar applies to the consultants in our network. All of them work as lead developers, architects and mentors for your staff, and the knowledge stays with you.

A small company. A large network.

Since 2002 Agnitio has built up a large network of experienced consultants. When an engagement calls for more or different specialists, we bring them in — with the same bar for competence and the same way of working. You get the closeness of a small company and the reach of a large one.

Per Hammar

Architect, mentor and lead developer — and AI champion.

Per has worked with a wide range of tasks in IT and works as an architect, mentor and lead developer. He is very sociable and enjoys giving courses and seminars. Per has particular expertise in large, terabyte-scale databases and has been engaged by Microsoft to help their customers with advanced assignments. Today he works on how AI changes software engineering, workflows in large organisations and product design — and takes that change from the individual developer to the whole organisation.

SOLID and DRY are principles he works with extensively and teaches in projects. Domain-driven design, TDD, CQRS and event sourcing in several systems; Scrum and Kanban, including the Scrum master role in some projects; migration of older architectures to modern ones using event-driven architecture; and extensive experience of building scalable systems. .NET since the beta in 2000.

His first computer was bought at the age of seven. In secondary school he sold his first game, which was used in the school's technology classes, and as a teenager he competed in programming across the Nordic countries. Per has also studied chemistry at KTH.

What he has done for customers · Contact us

  • over 1.7 million

    lines of code in our own system where AI generates, refactors, tests and modernises — with us at the helm. Several million lines in total.

  • 100,000

    CVs indexed with Azure OpenAI embeddings for semantic search and matching.

  • 26

    AI agents of our own working around the clock — planning towards goals, e-mail, market analysis, MVPs, research, monitoring.

AI in practice, right now

The question has changed. No longer "how fast can we write the code?" but "how much of the problem can we specify, delegate to AI, verify and integrate?"

  • AI-assisted development

    Over 1.7 million lines in our own system, several million in total — and AI throughout the workflow.

    Not toy projects but our own, genuinely large system — a quarter of the lines are tests. The latest coding models generate code, refactor, fix compiler errors and failing tests, write and repair unit tests, investigate bugs, reason about architecture, modernise old code, explain unfamiliar parts of the codebase and accelerate the migration to newer .NET. We measure quality and defect rates against traditional development. The step is clear: from developer, via AI-assisted developer, to a developer orchestrating coding agents that take on ever larger tasks on their own.

    C#, .NET, the latest coding agents, unit and integration tests

  • MyPM — our own product

    100,000 CVs a language model can search — and AI in the everyday work of a staffing company.

    The CVs are indexed with Azure OpenAI embeddings for semantic search and candidate matching. Next is AI-driven scheduling of staff and assignments — a long-term competitive advantage — and automation of staffing-company administration, with the EU AI Act taken into account for recruitment and interviews. AI is also used in refactoring the architecture, including the event-sourced systems, and in integrations, payroll and accounting, ROT/RUT and AGI.

    Azure OpenAI, embeddings, event sourcing, .NET, the Swedish payroll and tax domain

  • 26 agents

    26 agents that work for us — around the clock.

    Per has built up 26 AI agents of his own. They are specialised and talk to each other: one orchestrates and plans towards goals, others organise the e-mail — sort, classify, archive and surface what needs action — watch the markets and write morning, evening and weekly reports, build MVPs, do research and competitor analyses, index documents and images, monitor systems and track prices. Every morning they deliver a briefing. They also improve themselves: constant and never-ending improvement (CANI) measured against a North Star metric, so that every week is better than the last. That is the difference between using AI and working AI-first: the human sets the tasks, reviews and integrates — the agents do the work in between. We bring the same way of working into your projects.

    Hermes, agentic workflows, coding agents, automation

  • AI champion

    From the individual developer to the whole organisation.

    Per is an AI champion at a large financial organisation and leads its internal AI Community of Practice — the work of taking AI from the individual developer to the whole organisation.

    AI adoption in a large organisation, Community of Practice

  • Under the hood

    We don't settle for using the models. We build them too.

    To understand how language models work we have implemented a miniature LLM from scratch in C# and explored 1-bit quantisation, ROM-like hardware implementations, activation arithmetic and accumulators, attention, the KV cache, normalisation and softmax, sampling, tokenisation and distillation training. That makes us credible across the whole chain: model, coding agent, architecture, business process and organisation.

    Transformer architecture, quantisation, distillation, C#

Course instructor since 2007

Instructor for Addskills, Cornerstone and Global Knowledge Sweden, teaching Microsoft's official SQL Server and C# courses: database administration and maintenance, implementing and designing the data tier, performance and query tuning, troubleshooting, high availability and upgrades — plus Programming in C#. Earlier, in-house courses on three-tier and scalable systems with MTS/COM+, and talks on aspect-oriented development.

Selected certifications

Certified IT Architect Master and a large number of Microsoft certifications — among them:

  • Certified IT Architect Master (DFK-CIA)
  • Microsoft Certified Trainer (MCT)
  • Microsoft Certified Solution Developer (MCSD)
  • Microsoft Certified Database Administrator (MCDBA)
  • Microsoft Certified Technology Specialist (MCTS: SQL)
  • Microsoft Certified IT Professional (MCITP: DB ADM DEV)