DSML.KZ Community Hub

The AI/ML community connecting people, knowledge, research, and opportunities across Kazakhstan and Central Asia.

Join community What is DSMLKZ

Jobs

Browse roles

Jobs

Find AI, ML, data, and IT roles shared with the community.

Research

Explore hub

Research

Discover researchers, projects, papers, and collaboration signals.

News

Read updates

News

Follow AI/ML updates, community news, events, and opportunities.

Members

Find people

Members

Find people learning, building, researching, and hiring in AI/ML.

Events

See events

Events

Join meetups, talks, reading clubs, workshops, and recordings.

Channels

Follow DSML across Telegram, YouTube, LinkedIn, and community feeds.

Your profile as

multidimensional

as your data

Research

Import your work from ORCID, share interests, and find collaborators

Community

Ask personal questions, join discussions and stay connected

Careers

Add your CV, share job preferences, and manage your applications

Community Feed

Nursultan Tolegenov@ntolegenov

New DSML profileOctober 2, 01:01

Astana, Kazakhstan

IT specialist with experience in cybersecurity and software development. Fast learner, technically minded, and motivated to solve problems and grow professionally.

Open profile

Frontend Developer (React + TypeScript) at Armeta AI

New vacancyOctober 1, 21:30

Astana / Office · 600,000 to 900,000 KZT NET per month

AI for the construction industry in Kazakhstan. A platform for checking design documentation, cost estimating and working with building codes that saves engineers and designers time. Responsibilities: • Build new screens and features in React/TypeScript and ship them to production. • Integrate with the REST API (OpenAPI client), maintain and refactor existing code. • Write tests (Vitest, Playwright), take part in code review, optimize performance. • Follow the project architecture (Feature-Sliced Design). Requirements: • 3+ years of commercial React development • strong TypeScript • React 18+ • TanStack Query • Zustand • Tailwind CSS • Vite • experience writing unit and e2e tests • confident with Git • self-driven • responsible • attentive to detail Optional: • large data tables • WebSocket/SSE • CI/CD (GitHub Actions)

Open vacancy

Codex не спит, Claude управляет приборами — защита буксует

AI Radar editionAugust 29, 18:31

16 NeurIPS 2026 contests

Community newsAugust 30, 15:08

NeurIPS 2026 is coming, with the Competition Track featuring 16 competitions on various topics: • Scientific AI, Physics & Engineering: Learn2Design, Smart Buildings, Fusion Equilibrium, RealPDE, FAIR Universe Weak Lensing. • Healthcare/Bio: SimulacraBench, AMP Challenge, Speech Accessibility Project Challenge 2, Virtual Embryo. • Foundation Models, Reasoning & AI Evaluation: Steerability, Predictive AI Evaluation, AIMO Interpretability, QuantiPhy. • Robotics, Agents & Embodied AI: RoboSyn, RoCo-Spring, Agenthon for qunatitative finance Several more competitions will take place as part of workshops: EEG/EMG Foundation Challenge, AV Causal Reasoning Retrieval Challenge, Lean Refactor Arena, AgentOdyssey. Most competitions run in September and October. It is a great opportunity for anyone interested—including school students and people outside research—to tackle frontier problems faced by top labs, publish a paper about their solution, and attend the largest AI conference.

2Read news

Daniel Dragonevskiy

New research profileSeptember 30, 23:37

Faculty of Computer Science, HSE University, Moscow / ML Engineer at AIRA · Muscat

My research sits at the intersection of ML, LLMs and computer vision, with a focus on inference optimization (KV-cache compression, quantization) and ensemble methods for large-scale forecasting. In my recent paper (IJCISIM 2026, DOI: 10.70917/ijcisim-2026-4333) I showed analytically and empirically that geometric log-space ensembling is the MAPE-optimal aggregator for log-normal targets, reaching MAPE ≈ 4.77% on an 18,000-store retail forecasting challenge. Open directions: efficient LLM inference, vision-language models, and transferring multiplicative-error ensembling to new domains