Key Data Science projects

Price Forecasting  [GitHub link]

Forecast of UK wholesale energy prices for the next 24 hour period to inform bidding strategies.

• Data acquisition via API calls to NESO (National Energy Systems Operator)

• Data cleaning, EDA, and addition of lag features to inform forecasting

• Implementation of LSTM model using TensorFlow. Optimisation using Optuna. Deployment on Azure

Optimisation framework for resource allocation

A constrained optimisation-based matching engine built in Python to match 800+ MSc students seeking their dissertation projects to 200+ academic supervisors.

• Created pipeline for data acquisition from multiple sources (operations, IT, HR), pooling data and handle mismatch, data cleaning, ensure robustness.

• Optimisation using multi-criteria inputs while managing operational constraints.

• Direct and significant positive impact on operational efficiency supporting a £24M+ annual revenue stream

• Due to confidentiality reasons, the source code can't be shared openly.

Model identification framework for structural mechanics [Github link]

One of the long-standing problems in the field of mechanics of materials and structures is correct identification of the constitutive models that describe mechanical behaviour of a given material. In the domain of hyperelasticity, constitutive model is in the form of an energy density function (a function that describes how much potential energy is stored in the material due to elastic deformation).

In this project, we applied the EUCLID framework to automatically discover the appropriate constitutive models for a material given the experimental data.

• Use of a novel input convex neural network framework such that the model automatically satisfies the laws of physics.

• Ensemble modelling and statistical validation to assess model robustness

• Formulation of Physics-informed loss functions

Scientific Computing:

During most of my scientific career, I have focussed on developing computational solutions to engineering problems. Below are some recent projects. I have published over 30 scientific papers which can be downloaded from here and can be seen on Google Scholar. 

Continuous learning:

I have completed the following certifications in Machine Learning and Data Science

• Transformer Models and BERT models [certificate link]

• Trading, Machine Learning and GCP [certificate link]

• Using Machine Learning in Trading and Finance [certificate link]

• Financial Markets, Yale [certificate link]

• Portfolio Construction and Analysis with Python, EDHEC business school [certificate link]

• Inferential Statistical Analysis with Python, Michigan [certificate link]

• Deep Learning with PyTorch: Generative Adversarial Network [certificate link]

• Machine Learning, Stanford Online [certificate link]