Skip to content

Ademi Zhanuzakova

@ademi-zhanuzakova

Machine Learning | Deep Learning | AI Agents

I like working on AI tools that feel useful in the moment. Recently I built an agent that joins a brainstorming board, and I also enjoy exploring how smaller models can run more efficiently.

Career

Experience: 1 years

Experience

Research Assistant

Nazarbayev University

A Kazakhstan-based research university in Astana, with undergraduate and graduate programs and research across science, engineering, and technology.

May 2026 — Present

Researches efficient fine-tuning for vision models, co-authored a preprint on GPU DVFS, and adapted pretrained models for blood pressure estimation from ECG and PPG.

• Researching parameter-efficient fine-tuning methods to optimize computer vision models based on Transformer and Mamba architectures for Edge AI tasks. • Co-authored a preprint on energy-efficient GPU frequency management (DVFS) during fine-tuning of small language models. • Adapted pretrained BENDR, FEMBA, and CEReBrO models to estimate blood pressure from ECG/PPG signals using transfer learning and fine-tuning.

Machine Learning Engineer Intern

Arlan Biotech

Jun 2024 — Aug 2024

Built a PyTorch pipeline to classify epitope paratope interactions from biological sequences, reaching an F1 score of 0.75.

• Developed a deep learning pipeline in PyTorch to classify epitope paratope interactions from biological sequences. • Applied data preprocessing, class imbalance methods, iterative experiments, and error analysis, improving model quality to F1 = 0.75. • Used ProtBert embeddings to extract Transformer features from amino acid sequences and train an ML model.

Machine Learning Engineer Intern

BeyimTech

Jun 2024 — Jul 2024

Built an adaptive testing system and a Transformer based pipeline to grade open ended answers using semantic similarity.

• Developed an adaptive assessment system in Python to personalize test difficulty to the user's level. • Set up semantic similarity matching between user and correct answers to assess meaning similarity. • Implemented an NLP pipeline to automatically grade open-ended answers using Transformer models.

Education

Bachelor's degree, Computer Science

Nazarbayev University

2023 — 2027

Selected projects

AI Agent for Brainstorming

Developed an AI agent for a collaborative canvas that takes part in brainstorming sessions directly on the board. Designed its prompt logic and API workflow, including goals, context handling, and limits on generated responses. Integrated multimodal AI model APIs to generate images and video on the interactive board.

Open project

AI Assistant for Meal Planning

Pet project mentioned as part of the Product Management course.

Skills

PythonSQLC++LLMsAI AgentsPrompt EngineeringPEFTTransfer LearningClassificationRegressionTransformersCNNsMamba/State Space ModelsFine-TuningREST APIsGitLinuxJupyter NotebookKaggle