I am a quantitative researcher, with an overall research experience of more than eight years, specialised in mathematical and numerical techniques, non-linear dynamics and stochastic processes. I am a physicist by basic training, and a mathematician (or problem solver) at heart. Given my new-found passion for artificial intelligence and data science for impactful real-world problem-solving, I am looking to reorient my career in this path.

--------------------- New interest and explorations ---------------------

Deep Learning

- Image classification using transfer learning

Developed a model to classify images of Vajpayee, Abdul Kalam and Sadhguru using transfer learning with mobbilenetV2 model. However, this model has no practical use; the purpose of this project is to understand how to collect data required to train the model using web scraper, employ transfer learning and deploy the model to cloud. The model can be tested at https://classify-image-demo.herokuapp.com/.

Quantitative Finance

- Portfolio optimization

The standard deviation of returns of a share is considered to be a measure of its risk. If we have several shares in a portfolio, is it possible to pick weights for the shares of the portfolio in such a way that the risk of the portfolio is smaller than risks of the individual shares? How to find the optimal portfolio which minimizes the risk while maximizing the returns of the portfolio? I will address these questions using Modern portfolio theory by considering a portfolio of ten companies from NSE, India.

- Capital Asset Pricing Model (CAPM)

work in progress

- Stock price prediction

One of the assumptions in a mathematical model of financial securities is that the future stock price is a random variable. If that assumption is reasonable, is it possible to predict a future stock price? How much accuracy can be achieved? I will answer these questions using various models such as Logistic regression, KNN, SVM and LSTM network. The direction of the stock price is predicted rather than the actual price of the stock in this analysis.

- Highly correlated Indian stocks

Stocks of similar companies behave similarly. Due to some unknown reason, if their relative prices diverge, we can expect their relative prices to converge at some time in future. When prices of similar stocks diverge, statistical arbitrage strategy (also called pairs trading) is employed to exploit the situation. How can we calculate similar stocks from the whole list of stocks? I considered top 100 Indian stocks by market capitalization and calculated their similarity (or correlations) for all the possible pairs.

Statistics

- Statistical analysis of NAA and Mi concentration levels in Cingulate gyrus

I did this project as part of the Mathematical and Statistical Methods course at IIT Bombay. In this project, I employed hypothesis testing and ANOVA to obtain relevant conclusions..

----------------- Background and accomplishments -----------------

Algorithms

- Creating analytically divergence-free velocity fields from grid-based data

Developed an advanced algorithm that creates a continuous analytically divergence-free vector field (or velocity field) from a discrete vector field and published the method here. This method is necessary to track fluid particle trajectories accurately.

- A novel numerical method to estimate joint density of states

Co-developed a novel numerical method to estimate joint density of states of a thermodynamic system using the Wang–Landau algorithm and published the method here

Stochastic Processes in Thermodynamics

- Fluctuations in Model Non-equilibrium Processes

Work required to drive a thermodynamic system from one equilibrium state to another equilibrium state through a non-equilibrium process (i.e. over a finite amount of time) fluctuates each time we experiment. The connection between the averages over an ensemble of non-equilibrium measurements (e.g. work) and equilibrium properties of the system is explored using Monte Carlo Simulations.

Nonlinear Dynamics and Chaos

- Lagrangian transport structures in 3D incompressible flows

Showed that global Lagrangian transport structures of a 3D time-periodic flow with one invariant can be completely understood and calculated numerically by identifying resonance bifurcation points that periodically have zero net deformation.

- Perturbation effects on Lagrangian transport structures via inertia

Found perturbed resonance structures and a new mechanism of 3D chaotic transport that features non-heteroclinic connections of tubular transition regions when inertia is introduced to the 3D time-periodic flow.

Journal Articles

2020. Bharath Ravu, Guy Metcalfe, Murray Rudman, Daniel R. Lester and Devang V. Khakhar. Global Organization of Three-Dimensional, Volume-Preserving Flows: Constraints, Degenerate Points, and Lagrangian Structure. Chaos: An Interdisciplinary Journal of Nonlinear Science, 30, 033124.

2016. Bharath Ravu, Murray Rudman, Guy Metcalfe, Daniel R. Lester and Devang V. Khakhar. Creating analytically divergence-free velocity fields from grid-based data. Journal of Computational Physics, 323, pp. 75-94.

2016. M Suman Kalyan, Bharath Ravu, VSS Sastry, and KPN Murthy. Joint Density of Stages Calculation Employing Wang-Landau Algorithm. Journal of Statistical Physics. Vol. 163(1), pp. 197-209.

Proceedings

2016. Murray Rudman, M.F.M. Speetjens and Bharath Ravu. The transition to fully 3D Chaos in Wavy Taylor Vortex flow. 20th Australian Fluid Mechanics Conference.

Ph.D.

· Indian Institute of Technology Bombay, India, and Monash University, Australia (a joint PhD)

Thesis: Lagrangian Transport Structures in Three-dimensional Incompressible Time-periodic Flow.

M.Tech. in Computational Technology

· University of Hyderabad

Thesis: Fluctuations in Model Non-equilibrium Processes.

M.Sc. in Physics

· University of Hyderabad

Thesis: Self healing properties of a caustic beam.

Data Science
Qunatitative Finance