Careers

Project Duration: 22 Weeks (Starting April 1, 2025, ending September 2, 2025)
Location: Remote

Project Overview
We are seeking a highly motivated Data Scientist Intern to join our AI Assist: ARP AI Data Project, which aims to develop a robust AI-based forecasting model for marine weather predictions. The intern will work on integrating data from multiple sources, including regional weather data, third-party global forecasts, and proprietary MarineLabs datasets. The primary focus will be on time series forecasting using advanced AI methods, including Graph Neural Networks (GNNs), to improve the accuracy and reliability of marine weather forecasts. Experience with computer vision model development for image classification and object detection would be an asset for secondary projects.

Project Objectives

1. Develop AI-Based Forecasting Models:
   • Build a functional AI-based forecasting model, leveraging GNNs for high-accuracy predictions of marine weather parameters (wind speed, wind direction, significant wave height and wave direction).
   • Target: Have a GNN up and running that predicts 10m elevation wind speeds, significant wave heights, and average wind/wave directions, for a 10-day forecast.

      

2. Improve Forecast Accuracy:
   • Achieve high levels of forecasting accuracy, with the goal of improving MarineLabs’ current 10-day forecast accuracy.
   • Achievable Target: Maintain forecast accuracy within 10% of normal levels even when some buoy data is missing.
   • Target: Achieve an improvement of at least 10% over current forecast accuracy, evaluated using Root Mean Square Error (RMSE) for a 10-day forecast.

      

3. Incorporate Additional Forecast Metrics:
   • In addition to wind and wave, investigate producing forecasts for additional metrics such as air temperature, pressure, relative humidity and probability of fog.
   • Define performance requirements for new forecasts and ensure new GNN model 10-day forecasts meet these requirements.

      

4. Data Integration and Robustness:
   • Integrate data from various sources, including real-time and historical data, to improve the forecasting model’s performance.
   • Ensure that the GNN model can function effectively even when certain sensor data is unavailable, maintaining forecast reliability under offline conditions.

      

5. AI Model Deployment:
   • Deploy the developed AI model within the project timeline, ensuring the model meets operational standards and can generate accurate marine weather forecasts.
   • Support the validation of model performance and accuracy against standardized historical datasets.

Key Responsibilities

• Model Development:
  • Research and develop machine learning models for time series forecasting, focusing on GNNs to improve the accuracy of marine weather predictions.
  • Develop architecture for a Multi-Modal GNN capable of processing and predicting from multiple data sources.
  • Develop unit-level test plans to verify model performance.

     

• Data Handling & Analysis:
  • Work with multiple data sources, including real-time weather data, sensor data, and third-party forecasts, to develop a comprehensive dataset for training the forecasting model.
  • Assess internal datasets for suitability in training AI models.
  • Sanitize and classify datasets to ensure consistency and quality for model training.

     

• Model Evaluation & Optimization:
  • Evaluate and optimize the performance of forecasting models based on predefined accuracy metrics (e.g., wind speed, wind direction, wave height and wave direction forecasts).
  • Execute system-level tests to validate model performance.
  • Tune model parameters as required to improve forecast accuracy.

     

• Data Gap Handling:
  • Develop methods to accommodate and quantify the effects of missing or erroneous data, ensuring the GNN model can maintain performance despite data gaps.
  • Ensure the model can operate robustly even when sensor data is missing.

     

• Testing and Validation:
  • Execute unit-level tests to verify model performance and validate results.
  • Evaluate the final model’s performance and produce a comparison report against existing forecasts.

     

• Integration & Documentation:
  • Integrate the final model into the operational workflow, ensuring its functionality within MarineLabs’ forecasting system.
  • Document the development process, methodologies, and results clearly, including comparisons to current forecasting models.

Key Deliverables

• A fully functional AI-based forecasting model capable of generating 10-day wind speed forecasts.
• A detailed validation report outlining the model’s performance against accuracy benchmarks (e.g., ±3.8 m/s for wind speed forecasts).
• Documentation of methodologies, results, and code for future use.
• Comparison report evaluating the performance of the final model against existing MarineLabs’ forecasts.
• A GNN model integrated into the operational forecasting workflow.

Required Skills and Qualifications

• Machine Learning Expertise:
  • Strong knowledge of machine learning algorithms, particularly in time series forecasting and Graph Neural Networks (GNNs).
  • Proficiency in Python, with experience in machine learning libraries such as TensorFlow, PyTorch, and Scikit-learn.
  • Experience with computer vision models for image classification and object detection

     

• Data Handling & Analysis:
  • Experience working with large datasets and performing data cleaning, pre-processing, and feature engineering.
  • Ability to integrate data from multiple sources (e.g., regional weather data, sensor data, global forecasts).

     

• AI & Forecasting Skills:
  • Experience with time series forecasting models and the challenges of predicting variables such as wind speed, wind direction, and wave direction.
  • Familiarity with anomaly detection methods and techniques for dealing with missing data or sensor failures.

     

• Academic Background:
  • Currently pursuing a Master’s or PhD in Data Science, Computer Science, Machine Learning, or a related field.

     

• Independent Problem-Solving:
  • Ability to work independently, tackle complex problems, and apply machine learning methods to real-world datasets.

     

• Clear Communication:
  • Strong communication skills for documenting work, explaining technical details, and collaborating with team members.

Preferred Qualifications

• Familiarity with marine data systems and weather forecasting processes.
• Experience with deploying machine learning models in real-world applications..
• Understanding of the challenges of forecasting in dynamic and noisy environments.

Support and Mentorship
This internship offers the opportunity to work alongside experienced professionals and contribute to a high-impact AI project. You will receive mentorship and guidance from the project team to ensure a successful learning experience and the completion of your contributions.

Salary: TBD