Harvard GSECurrently at Harvard Graduate School of Education

Hi, I'm Alma.
I build & study tools for learning.

Between human learning, machine learning, and the design that connects them.

Product manager and researcher asking one question on loop: How might we build AI systems that better understand, support, and augment human learning and cognition?

01 — Now

About me.

I grew up in Kosovo flag Kosovo, where the internet was my lab, my library, and my classroom rolled into one. That early self-directed, relentless curiosity is still the engine behind everything I do.

Today I'm getting my Ed.M. at Harvard's Learning, Design, Innovation and Technology track while leading product for AI-driven classroom tools at Elite Education. I care about the messy middle: where research on how minds learn meets the design choices and machine systems that shape what actually happens in a classroom.

I build

Shipping AI products that survive contact with real learners — so the system and the user improve together.

I research

Measuring what's actually happening — attention, memory, participation, and learning.

I advise

Helping teams design curriculum, policy, and tools that fit how minds actually learn.

02 — Work

Selected work.

Product

An AI co-pilot for live classroom discussions

Teachers track who's speaking, who's quiet, and which ideas land — without a dashboard that pulls them out of the room. Early pilots show lower cognitive load, more voluntary participation from quiet students, and teachers noticing patterns they used to miss.

Product StrategyAI / NLPLearning Design
Research

Stigma vs. support in mental-health online discourse

Compared public posts about depression, OCD, and ADHD against evidence-informed descriptions. Found peer support and distortion sitting side by side — indistinguishable to readers. Started as an ML project, ended as a measurement problem.

NLPEducational PsychologyMeasurement
Writing & Curriculum

County AI Literacy Curriculum Development

Designed an end-to-end AI literacy curriculum adopted across a county government — establishing a shared baseline for 4,000+ employees.

Curriculum DesignEssaysLearning Science
Side quest

UN Document AI Translator

Fine-tuned MarianMT on the UN Parallel Corpora and shipped a translator workflow that hits 93%+ cosine similarity to human translation — cutting the busywork around multilingual policy work.

MarianMTHugging FaceFastAPI

03 — Path

Backtracing.

  1. Harvard University

    → 2026

    Harvard GSE · Learning, Design, Innovation, and Technology

    Stronger grounding in learning science so design choices are defensible, not just intuitive.
  2. Elite Education

    2022 — now

    Elite Education

    Product/Project Manager.

    Led and strategized the development of 5+ EdTech products — from learner tools to classroom tools to internal mentoring and curriculum management systems — and at times, developed curriculum.
  3. University of California San Francisco

    2022

    University of California San Francisco

    Junior Data Scientist / Research Assistant.

    Learned to work with vast amount of genomic data, and translate complex data insights for non-technical teams.
  4. University of Chicago

    2021

    UChicago · Data Science Institute

    Junior Data Scientist / Research Assistant.

    NLP on 40k+ tweets. Stepping stone to measuring meaning at scale in an academic research environment.
  5. Minerva University

    Research Assistant (Educational Psychology).

    Exploring nuance in the learning of science and how technology aids learning. Also a Teaching Assistant for Theories of Cognition and Emotion.
  6. Minerva University

    2018 — '22

    Minerva University

    B.Sc. Computer Science & Social Sciences. Got obsessed with how people learn — and how machines learn, too.
  7. Respect Our Rights

    Then

    Respect Our Rights

    Founding member of ROR, a nation-wide, child-led organization in Kosovo advocating for children's rights.

04 — Thinking

Principles I keep coming back to.

On effort

Learning that is hard is the learning that sticks.

On metacognition

Self-regulated learning is one of the biggest ROI skills.

On collaboration and society

Learning is almost never fully individual.

On priorities

Learning goals first, learning tools next.

On design

Tech and learning can work together miraculously — but good design is very needed.

I also write longer thoughts on Substack→

05 — CONTACT

Working on something at the edge of learning & AI?

I'd love to hear about it. I'm always up for thoughtful conversations, research collabs, or a good cup of coffee in Cambridge.