Middle Data Scientist

We are looking for a Middle Data Scientist with strong hands-on experience in time-series, deep learning, and Python development.
This role is a good fit for someone with 3+ years of professional experience in Data Science or Machine Learning, a strong understanding of neural networks, and practical experience developing predictive models for time-series and tabular data.
We expect solid knowledge of different neural network architectures, including recurrent and convolutional networks, as well as practical experience designing, training, tuning, validating, and deploying models.
Strong Python programming skills are essential for this position. You should also be confident working with statistics, data analysis, and production-oriented Data Science code.

What you will work on:
Perform exploratory data analysis and prepare statistical summaries for hypothesis testing and data-driven decision-making.
Build data preparation and feature engineering pipelines used by predictive models.
Design and implement predictive modelling approaches.
Continuously analyse and improve models already running in production.
Take an active role in Python development related to end-to-end ML model delivery and support internal ML-related libraries.
Tune loss functions, evaluation metrics, sample weights, hyperparameters, and other model components to improve model quality.
Prepare detailed model evaluation reports, summarise findings, and communicate practical insights to stakeholders.

What we expect from you:
At least 3 years of professional experience working as a Data Scientist or Machine Learning Engineer.
Practical experience building custom neural networks with PyTorch.
Strong knowledge of neural network architectures applicable to time-series forecasting and predictive modelling of tabular data.
Experience working with layers such as LSTM, GRU, Conv1D, Dense, Dropout, and BatchNormalization, as well as activation functions including ReLU, Sigmoid, Tanh, and Softmax.
Hands-on experience working with time-series datasets, including data loading, cleaning, preparation, and feature engineering with pandas and NumPy.
Experience with the Darts library would be a significant advantage.
Experience selecting and developing models available in Darts, as well as creating custom neural network architectures.
Excellent Python knowledge and confidence working with libraries such as Scikit-learn, Pandas, PyTorch, and SHAP.

Python proficiency is particularly important for this role. Our projects require Data Scientists who can independently write, structure, and maintain Data Science functionality and supporting libraries.

You should also have:
An understanding of statistics and data analysis.
Experience selecting and tuning loss functions such as MSE, MAE, Cross-Entropy Loss, as well as custom losses suitable for forecasting problems.
Experience with hyperparameter optimisation tools such as Optuna or Hyperopt.
Ability to implement and optimise PyTorch training loops, including data loaders and training/validation stages.
Knowledge of cross-validation approaches designed specifically for time-series data.
Understanding of regularisation techniques used to reduce overfitting.
Strong knowledge of model evaluation metrics, including MSE, RMSE, MAE, MAPE, and R-squared.
Familiarity with backtesting approaches used to validate model performance.
Experience with model deployment, including model serialisation and deployment for real-time services.
Experience preparing visualisations with matplotlib, seaborn, or Plotly.
Confident use of Git for version control and team collaboration.
Strong self-management and organisational skills.

What We Offer:
Competitive remuneration package.
Bonuses
Professional mentorship and guidance from experienced team members.
Opportunities for professional growth and continuous learning.
A dynamic and collaborative work environment.