Profile Pic

LinkedIn GitHub Email Me Resume

Welcome to Prithwish Maiti’s Portfolio

Welcome to my webpage.

I work at the intersection of systems engineering and data engineering, applying analytics to enable machines to perform high-level tasks with the computing power available today. My experience spans domains such as embedded systems, financial services, and healthcare analytics, where I have developed a deep appreciation for the need to design robust, scalable systems capable of processing data at scale.

I have an interest in dissecting complexes financial assets and leveraging the power of data to find profitable trends in them. I am eager to deepen my expertise in modelling information flow and reinforcement learning. With the vast amount of data available in today’s world, integrating data science models has become an essential factor for driving informed decision-making, optimizing processes, and unlocking new opportunities for innovation.

Highlights

Professional Summary

In my current role, I’ve focused on building low-level software systems as a part of Operating systems for portable communication devices. I’ve developed service-level functionalities with clean, user-friendly APIs, and engineered secure middleware for database communication, quantum-safe encryption, and video codecs. I apply best practices in memory management, resource allocation, and object-oriented design, while orchestrating concurrency with thread pooling and synchronization to deliver latency-critical solutions.

Two achievements I’m most proud of are: first, creating a live object detection service from camera feeds using TensorRT deserialization and high-speed streaming; and second, architecting an autonomous UAV system with state-machine execution, advanced features like live object and RF signal detection. These projects really demonstrate my ability to take complex problems and break them down to low-level problems and deliver optimized, production-ready end-to-end systems.

In my previous role, I worked in a financial services company, specializing in brokerage services and market intelligence of equity derivatives. I worked on a project on removing noise from limit order book by filtering out spurious orders (those posted by high frequency traders, not containing any market information). The purpose of this project was to improve the signal provided by trading balance to drive the market returns. While I was primarily working on a independent python analytics pipeline, I also got to explore the companies tick-by-tick data processing architecture following a subscriber-publisher C++ framework

My first role was data scientist in a healthcare analytics company. My work was on data validation. The tech stack included for programming data processing pipelines, Hive SQL for database and Hadoop for cluster management.

With my knowledge of statistics, data science, and programming, I am positioning myself to work at conjunction of quantitative financial analyst and developer.

Academics

I have a Bachelors degree in Mathematics and Computing from Indian Institute of Technology, Delhi. I secured a All India Rank of 280 in JEE Advanced 2017 to get an admisssion into this prestigious institution.

Course Done - Computer Vision, Deep Learning, Data Structures and Algorithms, Data Mining, Database Systems, Operating Systems, Theory of Computation, Optimization Theory, Statistics, Functional Analysis, Probability, Algebra, Differential Equations

I also have a Masters in Financial Mathematics from North Carolina State University, Raleigh, NC.

Course Done - Financial Data Science, Stochastic Calculus, Monte-Carlo methods, Financial Risk Analysis, Applied Bayesian Analysis, Fixed income Analysis, Statistical Learning

Research Interests

My current focus is on leveraging the power of large data sets to build statistical models for low-risk investment frameworks, such as statistical arbitrage. Additionally, I am interested in using computational power to achieve results in stochastic processes that were previously unattainable, such as pricing options. The courses in my financial mathematics program have provided me with the ideal launchpad for this work. I have a deep interest in efficient data structures and algorithms. I frequently engaged in coding challenges on platforms like LeetCode, CodeChef, and Codeforces, which I found to be a fun pastime. During my undergraduate studies in mathematics, I took courses in algebra, probability, PDE, ODE, and discrete mathematics in the first two years, before shifting my focus to data science, including a course in deep learning. My academic projects reflect my passion for pushing the boundaries of what’s possible with data.

I love developing alphas in my free time and post my research and result in github. But lately, I have not got the time to develop alphas, due to the full-time employment I am perusing. I feel that it the statistical quantitative part of a financial analysis that I enjoy the most.

Personal Insights

My approach to problem-solving is deeply analytical, combining theoretical knowledge with practical applications. I enjoy breaking down complex problems into manageable tasks and finding elegant, efficient solutions.

Prior to joining Murano Corporation, I had no experience of system-level design or low level programming. I started from square zero and went on to master every service, API, and concept our software uses. This experience has given me the belief that that I can tackle any problem that I determine to tackle. I also learned to emphasis soft skills like ownership of project, quality of work and more generally being responsible, which has added to my personal development. I learned how to become a valuable member of an enterprise.

Learning finance has transformed my perspective. Through the financial mathematics course during my bachelor’s program, I became fascinated by the mechanics of money and its true value. It gave me a deep, intuitive understanding, particularly in areas like options and futures pricing. I have been in other industries, producing analytical results and software development but didn’t feel the thrill of perusing quantitative research in financial Industry. But said that, a lot of the skills needed in those industries are transferable. Ultimately from my understanding, a person becomes valuable by developing niche skills and the ability to learn fast and accurately.

At IIT Delhi, I was surrounded by some of the best and brightest thinkers in the country, which fostered a competitive mindset in me. However, I also recognized that the education system often promotes evaluation-based learning, which can hinder a deeper understanding of subjects and contribute to a sense of a “rat race.” I am fortunate that I navigated that phase of my life and had the privilege to be mentored by amazing well-wishers.

Projects

Robust Risk Measurment and Worst Case Senarios

Financial risk measurement depends on models of prices and other market variables, but these models are inherently based on imperfect assumptions and estimates, leading to model risk. Optimization decisions, such as portfolio selection, exacerbate the impact of model error. In this study, we develop a framework to quantify the impact of model error and to measure and minimize risk in a manner that is robust to such errors. Our robust approach begins with a baseline model and identifies the worst-case error in risk measurement that could result from deviating from this baseline model, subject to a specific constraint on the plausibility of the deviation. By using relative entropy to constrain model distance, we can explicitly characterize worst-case model errors. This method goes beyond simply addressing errors in parameter estimates to consider errors in the underlying stochastic assumptions of the model, identifying the greatest vulnerabilities to error. We apply this robust framework to issues of portfolio risk measurement, credit risk, and delta hedging.

Here is a link to the Review paper of the theory.

Here is a link to the Presentation.

Mortgage loans Severity Analysis

Fannie Mae single-family mortgage loans data contains a lot of information on the loans they have acquired. It includes a robust comprehension of the performance of the loans and any delinquency procedure if conducted. However much of the data is in crude form, that is reflects only the procedure in which it was collected and compiled. There is a need to refine the data and extract useful features. Here I present the procedure, that I have formulated to do the same and them form a predictive model to form a LGD prediction table and subsequently a severity model.

This data set presents a challenge of developing a prediction model. I have attached a report of the data with the prediction model to be updated later along with the Github repo. [Mortgage loans severity Analysis] (https://drive.google.com/file/d/1TBUfwxoCxNq1n0x-OVJUoyirST4WQzd-/view?usp=sharing)

Algorithmic Trading Development

Currently, I’m working on deploying a systematic algorithmic trading strategy which mostly works on technical analysis. My main aim is to integrate data science logic to direct patterns using time series data into the trading strategies.

I am researching On 3 python libraries backed Trader back testing and pyAlgotrade(currently archived, so will move on to basana). Out of these three I have made significant progress in the pyAlgotrade library. Please stay updated for upcoming results which I will post on my website and my github repository.

Check out my blog on comparison of 3 algorithmic trading framework available as a python module

Blog

Market making Insights

Technical Articles

Contact

Get in Touch

I’m always open to new opportunities and collaborations. Feel free to reach out to me via email or connect with me on LinkedIn.

© [Prithwish Maiti] 2024 Home About Me Projects Blog Contact