I design conversational AI, build it, and put it in front of real users to find out what actually works.

Doctoral researcher at KTH Royal Institute of Technology in Stockholm, working where conversation design, LLM systems, and human–robot interaction meet. Three of my systems have gone into live use — an adult-education classroom, a university course, a middle school — and each one produced a paper about what it taught us. More at ali.mk.

Open to work — available immediately

I'm looking for roles in conversational AI, dialogue and interaction design, or applied research. Stockholm or remote from Sweden. I hold permanent residence, so there's no sponsorship or permit lead time. My thesis is due December 2026 and I can start alongside it.

A Swedish-conversation practice platform, deployed in an SFI adult-education school in Stockholm. I wrote and won the 2,000,000 SEK Digital Futures grant that funded it, then led the build:

• The product — React/TypeScript front end, Flask REST back end (~35 endpoints), JWT auth, and a multi-tenant schools/classes/students model with separate administrator, teacher, and student dashboards. Deployed under Docker and nginx with TLS.
• The conversation — Gemini and GPT-4.5 generate the dialogue; a Furhat social robot is the physical interface, bridged to the platform over Socket.IO so a teacher can provision a class and start a live robot session from the browser.
• The judgement — a separate LLM-as-judge stage in the live pipeline decides which corrective-feedback strategy to apply when a learner makes a mistake. Its behaviour is grounded in our published human-rater comparison of immediate vs. delayed feedback rather than in guesswork.
• The rollout — I persuaded the school to pilot it, ran the pilot across several teachers and classes of adult migrant learners, and delivered the onboarding workshops for the teachers, the students, and the principal.

The repository is private because real learner data runs through it. I'm glad to walk through the architecture and the design decisions.

Seven papers on how people learn language through conversation with machines, at HRI, IVA, CogSci, and in Frontiers in Education — plus the Sustainability Award at HRI '25 and an ACM SIGAI travel grant. Recent:

• Personalized Language Learning with an LLM Chatbot: Effects of Immediate vs. Delayed Corrective Feedback — Frontiers in Education, 2026 · 10.3389/feduc.2026.1703664
• Comparing Monolingual and Bilingual Social Robots as Conversational Practice Companions — HRI '25 · 10.1109/HRI61500.2025.10973901
• Conformity and Trust in Multi-party vs. Individual Human-Robot Interaction — IVA '24 · 10.1145/3652988.3673954

I review for IEEE RA-L, ACM THRI, and Advanced Robotics, and I've supervised 11 thesis students across KTH, Trento, and Freiburg.

Conversational AI — dialogue design, prompt engineering, LLM-as-judge evaluation, agent orchestration, local models (Ollama, LM Studio) · Build — Python, TypeScript/React, Flask, Kotlin (FurhatOS), Docker, Socket.IO, Git · Evidence — experimental design, participatory design, user research, mixed-effects models in R (lme4), Pandas/SciPy · Languages — Persian and Azerbaijani (native), English and Italian (fluent), Swedish and Russian (elementary)

Before KTH I was an Associate Researcher at Furhat Robotics and a research assistant at Tilburg University, after an M.Sc. in Cognitive Science at CIMeC, University of Trento. I like open source, and I build the tools I need rather than waiting for them.

ali.mk · LinkedIn · ORCID · alimk@kth.se