About Me

Hi, I am an Information Technology engineer learning my trade in applying mathematical and data science techniques to solve challenging problems in the domain of finance.

The applications of analyzing data model behavior and use it to make decisions interest me. I am currently working on dynamic portfolio management using deep reinforcement learning techniques.

Download Resume

2019

Lorem ipsum dolor sit amet, quo ei simul congue exerci, ad nec admodum perfecto mnesarchum, vim ea mazim fierent detracto. Ea quis iuvaret expetendis his, te elit voluptua dignissim per, habeo iusto primis ea eam.

2018

Strategy Research & Backtesting Intern, Parekh Shares and Securities
Researched and studied low frequency algorithmic trading strategies based on technical indicators. Worked with the backtesting team to develop a multistep backtesting strategy.

2017

Data Science Intern, University of Tor Vergata
Worked under Dr. Mauro De Sanctis to devise a methodology to analyze big datasets using Data Mining Algorithms under time constrained environment.

2016

Trading & Equity Research Intern, Parekh Shares and Securities
Worked with equity research team and used technical analysis and candlestick charting techniques. Executed client trades and made recommendations to them based on the research.

2015

Shop Floor Intern, Fabtech International Pvt Ltd.
Studied the designing, prototyping and assembly of Fluid Bed Processor, Closed Loop Granulizer and Single Pot Processor used in manufacturing of drugs.

2014

Took my first steps towards a career in mathematics as I led a three member team to the second place in Infinity Research competition held by Bits Pilani. We worked on Godel's theorem and outlined its applications in the observable environment.

What I Do

Mathematical Modelling

I work with modeling continuous and discrete time processes and interest in modeling complex financial instruments like exotic derivatives products.

AI Solutions

I am interested in the domain of imparting cognitive intelligence to systems to optimise and enhance their functionality.

Data Analytics

I am skilled at applying sophisticated visualisations and advanced statistical tools to generate inferences for smarter decision making.

NLP

I have worked with designing processing pipelines and performing sentiment analysis for web articles.

Computer Vision

I have worked on problems like OCR,handwiring, object and face recognition and am capable to solving problems in this domain.

Web Design

I have worked on developing static and dynamic webpages both on the front and backend.

Skills

85%

C++

Adept at using STL, developing Object Oriented and Structured applications.

75%

Python

Proficient at performing data analytics and developing machine learning applications.

70%

Java

Capable of developing core Java applications, graphical interfaces and website backend functionalities.

60%

R

Proficient at performing data analytics and developing machine learning applications.

65%

SQL

Capable of performing simple and advanced queries, triggers, procedures, functions and query optimisations.

55%

Matlab

Capable of logical programming and working with the image processing, neural networks, fuzzy logic toolboxes.

Projects

  • All
  • Finance
  • ML
  • Others

Statistical Arbitrage based Trading Strategy

Finance

Developed a pairs trading strategy by implementing the Engle Granger procedure to test for cointegration. Fitted the resultant spread between pairs to an Ornstein Uhlenbeck process to find a trading strategy and generate buy/sell signals.

Employed Kalman Filters to compute adaptive cointegration weights and OU process parameters to tackle loss of robustness.

Developed an algorithm to simulate backtesting for the strategy, implemented 18 performance metrics and used PyFolio tearsheet to analyze the structure of P&L from the strategies.

The handglider game

Others

Lorem ipsum dolor sit amet, consectetur adipisicing elit. Nulla, ratione ullam dolorum, officia quo illum asperiores incidunt ab quod repellat velit laudantium, quos. Earum eius dicta optio commodi repellat, mollitia.

Lorem ipsum dolor sit amet, consectetur adipisicing elit. Adipisci maxime mollitia cumque, dolore itaque ipsam nostrum ratione possimus architecto facere, aliquid eos voluptate id. Ea eum maxime, laborum expedita consequatur.

Self Driving Smartcab

Reinforcement Learning

Developed a self learning cab in a city environment which is capable of navigating from one point in space to another within a fixed acceptable time and with a high safety rating.

Used Q-Learning to develop the learning algorithm for the car.

Pricing a Synthetic Credit Portfolio

Finance

Priced a kth-to-default Basket CDS for 5 securities as an expectation over the joint distribution of their default times.

Computed default times for each reference from co-dependent uniform variables sampled from a Gaussian and Student-t copula. Calibrated the copulas by the correlation of equity prices for each name.

Computed the fair spread from discounted premium and default leg value for 1st to 5th to default instruments.

Adaptive Portfolio Manager(Others)

Portfolio Optimisation, Deep Reinforcement Learning

Developed a dynamic investment and portfolio management system for novice investors which optimises over user defined parameters like target return, risk aversion, asset preference, basket size and maturity.

The investment system first shortlists a set of investment opportunities from a universe of asset classes available to it based on user defined parameters. This project uses NLP and deep learning in the shortlisting proces.

Dynamic Portfolio Optimisation is performed over the selected set of assets. Deep Q-Learning is used for this optimisation.

Analyzing Customer Segments

Data Analytics & Unsupervised Learning

Analysed purchase patterns of customers for different items from multiple stores to analyse the impact in revenue and volume of change in supply chain method for a wholesaler.

Clustered purchase patterns to find segments within the data. Determine the respresentative behavior for each segment and then performed A/B testing to analyse the impact of any changes.

Interest Rate Modeling for CVA charge calculation

Finance

Computed CVA charge as specified within the Basel III accords for an Interest Rate Swap. The expected exposure on the contract is computed by simulating multiple interest rate curves using an HJM Framework based model.

Determined the number of factors in the model by performing PCA analysis on the yield curve. Calibrated the diffusion of the model from the principle components and the drift term was calibrated from the diffusion to satisfy no-arbitrage condition.

Dog Breed Classification

Deep Learning

Developed an algorithm to predict the breed of a dog from its raw image. If the input does is not of a dog, the model recognises the same and then outputs the breed with characterists closest to that in the image.

Used transfer learning and a deep convolutional neural network to develop the classification model. Introduced shear, magnification, translation, skew and similar image pre-processing techniques to improve the performance of the model.

Stock Price Prediction using LSTM's and ICA

Deep Learning

Designed and trained a Recurrent Neural Network model to predict the short term stock prices. Tested its performace against a hold and wait strategy and seld-designed metrics to test the optimism and pessimism of the model.

Computed 45 technical analysis based attributes from the tick data, performed ICA to reduce the dataset to 12 attributes and then run a LSTM and GRU model to predict stock prices from it.

30

Awards Won

310

Projects

4,710

Hours of Work

1,012,030

Lines of Code

Publications

Application of Swarm Intelligence to Portfolio Optimization

Meta-heuristic Optimization Akhil 28 Sept, 2018

Compared the feasibility, convergence and robustness of popular meta-heuristic optimisation algorithms for the portfolio optimisation problem.

Data Augmentation using Generative models for Credit Card Fraud Detection.

GAN, Credit Card Fraud Akhil & Raj 15 Dec, 2018

Tackled the class imbalanced problem for credit card fraud detection. Compared and Implemented Generative Adversarial Network and its variations to improve classifier performance.

Contact

If you wish to brainstorm over ideas and collaborate on interesting projects to solve challenging problems, please get in touch with me. Also, I would love to work on projects in Finance and Artificial Intelligence.

Copyright 2019 © Akhil Sethia. Developed With Akhil.