I build software that makes complex work simpler.

I’m Hossein Erfani. I build AI applications, connect systems through APIs, and automate complex workflows in higher education.

Explore selected work ↘

Selected work

Four systems, built around real needs in higher education.

01 / AI in education

KAU AI Tutor

AI support grounded in the course. Available inside Canvas.

KAU AI Tutor — instructor setup and course configuration.

Individual academic support is difficult to provide at scale. KAU AI Tutor brings AI support into the context of a course, where teachers and students already work.

My role. I designed and built the application end-to-end, including instructor configuration, student conversations and the integrations behind them.

Technical approach. A custom WordPress plugin on KIM, integrated with Canvas through LTI 1.3 and APIs, with RAG for course context and Berget’s language-model services.

Live in several courses. Course activation combines automated processing with final human approval.

View case study →

A longstanding interest: my master’s research explored student modelling in Intelligent Tutoring Systems. More about that background.

02 / Data integration & exploration

Course Evaluation Visualizer

From separate PDF reports to information people can explore.

Course Evaluation Visualizer — portfolio-safe interface view.

Course evaluations were primarily available as PDF reports. Comparing courses, following changes across course iterations, and exploring free-text responses across evaluations was difficult.

My role. I built the technical system: data integration, storage, synchronization and application functionality. A colleague created the UX and design.

Technical approach. The application retrieves data from Survey&Report, stores it in a WordPress database on KIT, and uses scheduled processes to keep it updated.

A dataset for comparison. The interactive dashboard supports exploring evaluations together and browsing their free-text responses.

03 / Integration & administrative tools

Programme Evaluation System (PES)

Connecting a workflow to the people and systems that operate it.

Karlstad University needed a digital system for programme evaluation across multiple roles and steps. The original Gravity Flow workflow was a team effort, with an external consultant contributing to its design and implementation.

My role. I extended the workflow with API integrations to Ladok and Canvas. I also designed and built administrative dashboards for status overviews, operational control and batch actions.

Technical approach. Integration functionality participates directly in Gravity Flow steps, bringing programme information, student lists and programme leaders into the process.

Used across approximately 200 university programmes.

Conceptual PES overview connecting Ladok and Canvas with evaluation stages, human roles, administration and reporting.

Programme Evaluation System — conceptual overview of roles, workflow and integrations.

04 / Process discovery & automation

Thesis Assessment System

Automating coordination while keeping academic decisions with people.

Conceptual thesis assessment workflow linking students, supervisors, co-assessors, examiners and administrators.

Thesis Assessment System — conceptual view of the role-based assessment workflow.

Thesis assessment depended on administrators coordinating emails, collecting information, preparing documents and moving each case between participants.

My role. I designed and built the system after mapping the existing process with administrators and participants. Their knowledge of the academic workflow informed the design.

Technical approach. Gravity Flow handles predictable coordination: notifications, data collection, document generation and conditional routing. Academic assessment remains with the appropriate people.

In refinement. Approximately 15 thesis cases in the first iteration exposed exceptions, some of which caused problems. The system is being refined to handle them better.

Understand the real process. Automate the predictable work. Keep people responsible for the decisions that need human judgement.

Presented Work

Selected projects I designed and developed that have been presented at conferences and professional events.

INTED 2026

Valencia, Spain

Conference paper & presentation

Customised Sovereign Chatbots Powered by OERs – Experiences from Sweden

J. Pareigis · H. Erfani

Project: KAU AI Tutor

This refereed proceedings paper describes Karlstad University’s digital-sovereign AI tutoring approach: LTI integration, WordPress as an extensible knowledge layer, OER-grounded content and institutional control.

View paper →

Sunetdagarna 2026

Norrköping, Sweden

Professional presentation

KAU AI Tutor

Project: KAU AI Tutor

The KAU AI Tutor work was presented at this professional event.

WordCamp Europe 2026

Kraków, Poland

Conference presentation

Sovereign university AI tutors powered by WordPress

Project: KAU AI Tutor

The session presented Karlstad University’s approach to institutionally controlled AI tutors using WordPress Multisite, open educational resources and university infrastructure.

View session →

TNC26

Helsinki, Finland

Conference presentation

Building lightweight digital-sovereign AI Chatbots for Higher Education: Connecting the dots of openness

Project: KAU AI Tutor

The contribution explored lightweight, digitally sovereign AI chatbots for higher education, connecting openness, institutional control and practical integration.

View official programme →

NU2026

Gothenburg, Sweden

Conference presentation

Thesis Assessment System

Project: Thesis Assessment System

The Thesis Assessment System was presented at NU2026 in Gothenburg, the Swedish national conference for higher-education pedagogical development.

A little background

I’ve been programming since 1993, when I first used a Commodore 64 in a classroom in Rasht, northern Iran. Discovering that I could give a machine instructions and make it do useful work set the direction.

I went on to study Software Engineering. My master’s research explored student modelling in Intelligent Tutoring Systems using Case-Based Reasoning. Through teaching, educational technology and software development, the recurring interest has stayed the same: making difficult or repetitive work simpler.

Contact

Let’s talk.

If you’re working on AI, automation, integrations or digital tools for education — or just have an interesting technical problem — feel free to get in touch.

Email →

LinkedIn →

Based in Karlstad, Sweden.