Skip to content
View t3i8m's full-sized avatar
:octocat:
Looking forward
:octocat:
Looking forward
3 followers · 9 following

Block or report t3i8m

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
t3i8m/README.md

Hi, I'm Timurs Jerčaks 👋

BSc Data Science & AI @ Maastricht University
Junior AI Engineer · ML Researcher · Builder

🧭 About Me

I'm a Data Science & AI student at Maastricht University with hands-on experience spanning financial analytics, NLP pipelines, and medical ML research. I enjoy building things that work - from CNNs from scratch to production-grade RAG systems.

  • 🔬 Thesis @ MUMC+ - Predicting Postoperative Delirium from intraoperative EEG signals
  • 💼 Junior AI Engineer @ LARKinfolab - GraphRAG architecture with ReAct agents
  • 🏦 Data Science Intern @ Luminor Group - Large-scale EDA (10M+ rows), built NLP extraction pipeline for 200K+ records, customer segmentation that reduced SMS costs

🚀 Projects

🤖 Machine Learning & Deep Learning

Project Description Stack
CNN from Scratch Convolutional neural network built with NumPy only - no PyTorch/TensorFlow. Trains on MNIST with custom conv, ReLU, softmax, backprop Python, NumPy
Multilayer Perceptron Handwritten digit recognition from a single image using a neural network built from scratch Python, NumPy
Simple Language Model Feed-forward neural language model in PyTorch. Fixed context window of 5 BERT token IDs predicting the next token on tinyshakespeare Python, PyTorch

🌌 Kaggle Competitions

Competition Score Approach
Stellar Classification (S6E6) 0.954 balanced accuracy XGBoost + Optuna + sklearn Pipeline + Error Analysis

📝 NLP

Project Description Stack
NLP Labs Labs covering Tokenization (BPE), Information Retrieval (TF-IDF), BERT, Prompt Engineering Python, HuggingFace
Comparative Study NMT Analysis of NMT architectures: Seq2Seq LSTM vs MarianMT vs M2M100 - accuracy, noise robustness, adaptability Python, PyTorch

🎮 Other

Project Description Stack
Yinsh game AI Yinsh board game with two AI bots: Alpha-Beta pruning and Deep Q-Network Java, JavaFX

🛠️ Stack

ML/AI
XGBoost LightGBM scikit-learn PyTorch HuggingFace LangChain optuna

LLM & RAG
ReAct Agents GraphRAG Neo4j Pydantic local models

Data
pandas numpy SQL matplotlib seaborn plotly

Engineering
Python Java FastAPI Docker Git Azure

📊 Currently Working On

  • 🧠 EEG-based ML pipeline for Postoperative Delirium prediction @ MUMC+
  • 📝 Writing about ML on Medium

📫 Connect

LinkedIn Medium Kaggle

"Train hard, generalize harder"

Pinned Loading

  1. CNN CNN Public

    Implementation of a convolutional neural network (CNN) from scratch using NumPy, without TensorFlow or PyTorch. Trains on the MNIST dataset with a custom convolution layer, ReLU, softmax, and backp…

    Python

  2. Multilayer-Perceptron Multilayer-Perceptron Public

    Handwritten digit recognition from a single image using a neural network

    Python

  3. nlp_labs nlp_labs Public

    NLP labs and reports | Topics: Tokenization (BPE), Information Retrieval (TF-IDF), BERT, Prompt Engineering and more.

    Jupyter Notebook 1

  4. Comparative-Study-NMT Comparative-Study-NMT Public

    A project for the comparative analysis of NMT architectures (Seq2Seq LSTM, MarianMT, M2M100), evaluating accuracy, adaptability, noise robustness and potential improvements.

    Jupyter Notebook

  5. Simple-Language-Model Simple-Language-Model Public

    A minimal feed‑forward neural language model built with PyTorch. It takes a fixed‑length context window of 5 BERT token IDs and predicts the next token. Using a tinyshakespeare.txt dataset

    Python

  6. yinsh-game-ai yinsh-game-ai Public

    This project is a Yinsh board game implemented with JavaFX with two types of bots: Alpha-beta prunning bot and Deep Q Network bot

    Java