Skip to content
AK

Atay Kim

@cerebrum

Applied mathematics | Scientific machine learning | Software engineering

Learning AI, DS/ML for solving physics simulation problems. Math enthusiast. Beginning researcher. Part of INMET modelling center. Part of Beyond Curriculum Public Foundation.

Open to

Hiring & referralsResearch collaborationHelping with community projectsPaper discussionsCompany partnerships

Career

Experience: 7 years

Experience

Research Assistant

INMET, Nazarbayev University

scientific machine learning for engineering PDEs

Jan 2026 — Present

Built and tested FNO benchmarks for heat conduction, measuring lower temperature and flux errors with more Fourier modes, and presented the work to about 180 attendees.

• Built a PyTorch Fourier Neural Operator benchmark for heterogeneous 1D heat conduction; generated 64-node finite-volume reference data and evaluated temperature, heat-flux, and discrete PDE-residual errors. • Designed a controlled comparison of near-equal 139k-parameter models using 4 versus 32 Fourier modes. In the hardest out-of-distribution regime, 32 modes reduced median temperature error from 1.170% to 0.709% and flux error from 17.062% to 11.204%.

Visiting Research Student

Ghent Analysis & PDE Center, Ghent University

Ghent University is a Belgian research university with an Analysis & PDE Center.

Jun 2026 — Jul 2026

Generated 300 smoke-flow simulations and trained an FNO, then evaluated rollout behavior and documented blur, boundary artifacts, drift, and distribution shift.

• Generated and standardized 300 synthetic 2D incompressible smoke-flow simulations with randomized sources and sinks using PhiFlow; implemented and trained a PyTorch FNO on an NVIDIA Tesla T4. • Evaluated autoregressive trajectories against numerical simulations and documented blur, periodic-boundary artifacts, rollout drift, and distribution shift; released reproducible generation, training, evaluation, and demo components.

Incoming Selected Participant

Q.AI × 01.AI AI Mentorship Program

Inaugural national cohort; Q.AI, Alem.ai Foundation, Astana Hub, and 01.AI

Oct 2026 — Present

Selected for the inaugural 20–25-person national AI mentorship cohort, which includes applied projects, workshops, and mentoring from 01.AI experts.

• Selected through a competitive national process for the inaugural 20–25-person cohort, assessed on English proficiency, AI experience, demonstrated projects, and motivation to build scalable AI solutions. • Program includes small-group applied AI projects, workshops on frontier AI and technology entrepreneurship, and mentoring sessions with 01.AI experts and Dr. Kai-Fu Lee, founder of 01.AI and Sinovation Ventures.

Education Consultant

BI Education (Aitanym)

Jan 2026 — Present

Created and checked 500+ math problems and 10+ adaptive lessons for an AI-personalized UNT preparation product, reviewing accuracy and lesson sequencing.

• Created and quality-checked 500+ mathematics problems and 10+ adaptive lessons for UNT preparation; translated exam requirements into structured content for an AI-personalized learning product. • Reviewed mathematical correctness, difficulty calibration, and lesson sequencing with the product team; revised long-form learning flows to reduce disengagement and poorly targeted practice.

Organizing Manager

Beyond Curriculum, Kazakhstan

Nov 2023 — Present

Organized five international contests for 2,000+ participants across four countries, coordinating teams of up to 130 and budgets over USD 20,000.

• Organized five international contests reaching 2,000+ participants from four countries; coordinated up to 130 organizers, jury members, designers, and volunteers and managed an event budget exceeding USD 20,000. • Led partner communication with Chevron, Air Astana, Haileybury, Nazarbayev Foundation, Al-Farabi Kazakh National University, Energo University, and Baiterek School.

Education

BSc, Mathematics, School of Computing and Artificial Intelligence

Nazarbayev University

2023 — 2028

Secondary education

National School of Physics and Mathematics

2021 — 2023

Selected projects

CoursePlanner

Built a degree-planning application with prerequisite graphs and semester-sequence validation, used by 2,000+ Nazarbayev University students per semester, approximately 30% of undergraduates.

Open project

ModalStripe

Investigated gated modal memory with local and global attention in small character language models; compared models at roughly 121,000 parameters over three optimization seeds and documented limitations in a research manuscript.

Open project

KZ RAG Eval

Adapted an evaluation harness to compare Qwen 2.5 3B with and without BM25 retrieval on 21 multilingual questions grounded in official Kazakhstan public-information pages across Kazakh, Russian, and English.

Open project

Fourier Neural Operator benchmark for heterogeneous 1D heat conduction

Built a benchmark using 64-node finite-volume reference data to evaluate temperature, heat-flux, and discrete PDE-residual errors.

Open project

Achievements

Silver Medal, International Youth Math Challenge

International Youth Math Challenge · 2022

Best result from Kazakhstan

International Youth Math Challenge · 2022

Absolute 2nd place, National Mathematics Olympiad of Kyrgyzstan

National Mathematics Olympiad of Kyrgyzstan · 2021

Full-tuition scholarship offer, USD 80,000+

City University of Hong Kong

Full-ride offer to Yıldız Technical University

Türkiye Bursları

1st place, AMC 10

AMC 10

International Zhautykov Olympiad participant

International Zhautykov Olympiad · 2021

International Zhautykov Olympiad participant

International Zhautykov Olympiad · 2022

Skills

PythonTypeScriptJavaScriptSQLHTML/CSSPyTorchNumPypandasMatplotlibPhiFlowfinite-volume simulationdata generationmodel trainingautoregressive evaluationstatistical experiment analysisReactReact FlowViteNext.jsNode.js