My Resume

You can find My Resume here 📄

Basics

  • Name: Andrii Nikolaiev, Ph.D. in C.S.
  • Status: Visiting researcher at the University of Cambridge
  • Email: adnikolaiev@gmail.com

Professional Experience

NLP Researcher at University of Cambridge

Apr 2023 – Present

  • Developed benchmark datasets and applied computational statistics, revealing up to 30% variance in LLM reasoning due to textual modifications.
  • Deployed language models (GPT, LLaMA, Mistral, Qwen) on HPC clusters and cloud VMs with llama.cpp.
  • Led human evaluation studies with 40 participants to validate benchmark difficulty level and assess model outputs against expert judgments.
  • Skills: Dataset Design, HPC Deployment, Model Benchmarking, Computational Statistics, Natural Language Processing.

Ph.D. Student at Taras Shevchenko National University of Kyiv

Oct 2020 – Sep 2025

  • Analysed and processed 860K+ Chain-of-Thought examples using Python and R.
  • Evaluated 15+ LLMs on automated reasoning using large-scale datasets.
  • Developed data generation methods and quality evaluation pipelines with LLM-as-a-Judge to automate math creation problems, reducing annotation costs by over 50%.
  • Skills: Foundation Models, Mathematical Reasoning, ATP, Data Augmentation.

Machine Learning & AI Projects

Historical Records Processing Historical Records Processing – Osiris AI
- Enhanced HTR pipeline accuracy by 10–15% on 19th-century tabular records.
- Integrated multimodal LLMs with PAGE XML (eScriptorium) and raw images.
- Deployed models with vLLM, Ollama on distributed VM infrastructure.
Solar Flare Classification Solar Flare Classification - Samsung ML Research
- Processed and visualised data, benchmarked traditional ML models.
- Trained neural networks for classification.
- Achieved 99% accuracy on the UCI Solar Flare dataset.
Automated Math Tutor Automated Math Tutor
- Prepared datasets by collecting and analysing real math problems and solutions from online classes.
- Curated design and implementation for LLM-based grading agentic system via Google Cloud Platform, achieving 80% accuracy.
- Investigated mathematical text segmentation and data extraction with Mistral OCR, Kraken.
Pandemics Forecasting COVID-19 Forecasting
- Compared SIR/SEIR vs. ML (Gradient Boosting, Random Forest, SVM, ARIMA) models on JHU CSSE dataset.
- Outperformed traditional models on short-term (1–2 weeks) regional forecasts.
NN engine Neural Network Games Engine
- Built Reinforcement Learning + Monte Carlo Tree Search engine for chess and Atari; no handcrafted heuristics.
- Used TensorFlow, PyTorch and Ray for distributed training on Google Cloud Platform.
- Reached 2000 ELO in 12 hrs.

Skills: Supervised/Unsupervised Learning, Deep Learning, Reinforcement Learning, Time-Series Forecasting, OCR, Data Analysis, mLLMs, MLOps.

Education

University of Cambridge, UK

  • Visiting Postgraduate Research Student, 2023 – 2025

Taras Shevchenko National University of Kyiv, Ukraine

  • Ph.D. in Computer Science, Oct 2020 – Sep 2025

  • B.Sc. & M.Sc. in Computer Science (with honours), 2014–2020

Additional courses

  • Machine Learning Specialisation, Stanford Online via Coursera (2022)
  • Genesys IT School (selected in top 1.6% of 2,500 applicants, 2021)

Grants & Awards

  • Two-time Grant Winner for innovative educational projects — Emergent Ventures U.S. fellowship from Mercatus Center at George Mason University (2023; 2024)
  • Award for 2-year fully funded visiting postgraduate research placement at the University of Cambridge (2023–2025)

Additional Experience

  • Leadership: Led STEM education programs (800+ students; $120K+ in grants)

  • Mentorship: Delivered 1,000+ hrs of instruction on Programming, Databases, Cloud Computing, and Mathematics

  • Public speaking: Delivered talks on AI and STEM at practical conferences and workshops

Skills

  • Programming: Python, R, MATLAB, C/C++, SQL
  • Frameworks & Libraries: TensorFlow, PyTorch, Ray, Hugging Face, llama.cpp, vLLM, Ollama, NumPy, Pandas, Seaborn, Scikit-learn
  • Deep Learning & LLMs: Transformers, LSTM, GPT, BERT, CNN, RNN, MoE
  • Cloud & DevOps: Google Cloud Platform, HPC clusters, Git, UNIX
  • Web & Markup: HTML, CSS, Markdown, LaTeX
  • Product: Market Research, Analytics, A/B Testing, Project Design
  • Interpersonal: Leadership, Entrepreneurship, Mentorship
  • Languages: English (fluent), Ukrainian (native)

Please, check out my Research and STEM Projects pages for more details.