Skip to content
VM

Vladislav Mun

@BoulderyBoulder

Senior AI/ML Engineer and Researcher

ML engineer with 6+ years of AI research experience and 10+ years of Computer Science. Interested in remote work or semi-remote work in Astana

Career

Experience: 6 years

Experience

Research Assistant · machine learning for drug discovery

Nazarbayev University — AI/ML Lab

Nazarbayev University is a research university in Kazakhstan.

Aug 2023 — Present

Led development and release of a 96M-parameter multitask transformer covering 683 biochemical endpoints; reported leading RMSE on three benchmarks.

• Designed and trained CheMLT-F, a dual-encoder DeBERTa model that fuses molecular (SMILES) and protein-sequence representations to serve 683 endpoints—toxicity and bioactivity classification, physicochemical property regression, and drug–target affinity—from one shared backbone (96M parameters, 34M trainable). • Pretrained both encoders from scratch across 6× RTX 4090: masked language modelling on ~58M PubChem molecules and 1.8M protein sequences, with byte-level BPE tokenizers trained on each corpus.

Research Assistant · explainable AI for neural decoding

Nazarbayev University — BCI Lab

Nazarbayev University is a research university in Kazakhstan.

Jun 2021 — Dec 2021

Applied explainable AI to EEG classifiers, tested relevance-guided model pruning, and built a Unity environment for BCI experiments.

• Applied layerwise relevance propagation to ERP/EEG classifiers, generating relevance heatmaps and scalp maps that identified which time windows and channels drove each decision. • Used LRP relevance scores to prune network pathways, quantifying how far model complexity could be reduced before predictive performance degraded. • Built a 3D BCI training and testing environment in Unity for motor-rehabilitation experiments.

Software Engineer Intern; Data Science Intern

Institute of Information and Computational Technologies

Jan 2019 — Aug 2019

Built research data visualization and collection tools, and prepared datasets for internal projects during two internships.

• Built a Unity-based 3D visualizer that let in-house researchers explore and sanity-check high-dimensional datasets interactively. • Developed an online markup tool for visual data collection as a full-stack web application (Angular front end, Django REST API), including query and rendering optimization. • Collected, preprocessed, and analyzed datasets for the institute's internal research projects.

Research Mentor (informal)

Nazarbayev University — AI/ML Lab

Nazarbayev University is a research university in Kazakhstan.

Present

Mentors two student groups working on polymer property prediction and explainable drug design.

• Advised two student project groups—polymer property prediction and explainable drug design—in biweekly meetings, helping them scope research questions, structure experiments, and set direction between sessions.

Teaching Assistant · Deep Learning; Algorithms

Nazarbayev University

Nazarbayev University is a research university in Kazakhstan.

Assisted teaching Deep Learning and Algorithms courses.

• Teaching Assistant for Deep Learning (Fall 2021, Spring 2022) and Algorithms (Spring 2021, Fall 2022).

Education

Ph.D., Computer Science

Nazarbayev University

2022 — 2027

M.S., Computer Science

Nazarbayev University

2020 — 2022

B.S., Computer Systems & Software

Kazakh-British Technical University

2016 — 2020

Selected projects

Fully local low-latency voice pipeline

Benchmarked open speech models on latency and output quality, then assembled selected components into a fully offline speech-to-text → LLM → text-to-speech loop. Served Gemma- and Qwen-family models locally under llama.cpp, tuning quantization and offload settings to fit each within a single GPU's memory budget while keeping end-to-end latency low enough for conversational turn-taking.

Reinforcement learning for game agents

Replaced scripted unit controller behaviour with a trained RL agent; built the environment interface, reward shaping, training loop, and evaluation against a human baseline for a platformer.

Skills

PyTorchHugging Face Transformersscikit-learnTransformer pretraining (MLM)Multitask learning and loss balancingMixed-precision and memory-efficient trainingExplainable AI (LRP)PythonCC++C#JavaCheminformaticsRDKitSMILESPubChemMoleculeNet/DeepChemKIBADavisDrug–target interaction