Quantifi - PL
platform that hosts quant tools
About PL and Quantifi
Established in 1944, Prabhudas Lilladher is a stock broking company in India with a vision of rising up to become India's leading financial services provider. It has been a journey of success, based on strong adherence to ethics, uncompromising dedication to quality and an attitude of excellence. Imbibed in our corporate culture, these founding principles have guided us to reach the pinnacle that we are at today. Over the years, PL has evolved from a standalone brokerage firm to a one-stop shop to companies for all financial services.
With a team of dedicated experts and a nationwide distribution network of branches, franchisees and associates, PL provides a comprehensive gamut of financial advisory services in the Institutional and Retail domain. Our range of services includes Equity & Derivative Broking, Investment Banking, Corporate Advisory, PMS, Online Trading, Loan against Shares, Mutual Funds, IPO's, Real Estate, Home Loan & Loan Against property.
Drawing on our team's profound technical expertise and years of extensive practical experience, we develop customized solutions that are unique to each of our clients, thus ensuring high levels of customer satisfaction.
Quantifi is a platform developed by the research team and myself. It hosts all the Quant tools developed by the company using Machine Learning algorithms, quantitative analysis, big data, data visualization and efficient data science techniques.
Quantifi.in
Quantifi.co.in a RESTful web application developed using Python and Django Framework that hosts over 50 proprietary Quant tools under various categories: Quant meters, Quant frameworks, Quant Research, Utilities & Dashboard, etc. I would like to highlight some of the best tools on it:
• Implemented a quant-based multi-asset portfolio management system, optimize it and build an investor portal which now has an AUM of ~ $45 million, increasing the revenue of the firm by 16%
• Developed a tool that helped automate Nifty 50, Nifty Midcap and Nifty Bank index values in the MSSQL database, created a prediction model for the three indices using supervised Machine Learning algorithms and developed a backtest module to improve accuracy of the prediction model
- Improved accuracy of the model by 33% than the previous model and saved time by improving efficiency by 375%
• Modelled the sentiment of the financial markets using data points from various international markets’ indices and 43 other data points to give historic and future intensities of making a market call: Hold, Buy and Sell
- Helped the company make intraday calls with an accuracy of 83%
•. Designed and implemented a tool that evaluates and predicts the direction of the financial markets' stocks and indices using a proprietary algorithm and optimized using ML algorithms – dimensionality reduction on 120 factors and logistic regression
• Programmed a Stock Grid tool which compares relative performance of stocks against indices to suggest which stocks to buy at what time of a crisis. This tool helped the company profit ~ $200,000 during the COVID-19 pandemic
• Performed back-test and performance analysis to improve accuracy by 20-85% across 7 tools I developed
• Help recruit and lead a team of 3 people who helped develop more quant tools and implement the tools I built on the cutting-edge mobile application - Prabhudas Lilladher App. Increased usage by 18% and downloads by 34% in 3 months.
Tools and Platforms
-
We used Python 3 to make the back-end tools.
-
Web Interface was developed using Python's Django Framework.
-
Website hosted on PL's Windows Server and configured using Apache Web Framework.
-
All of the companies' new and historic data was stored in a cloud-based Microsoft SQL Database.
-
Automated tools ran on an Ubuntu Server using Crontab.
-
The integration on PL India app was done using Android Studio and XCode.
-
Used the Accord and NSE Live data APIs.
-
Used plotly for graphs and data visualization.
-
Used Machine Learning algorithms with the help of PyTorch, TensorFlow and Scikit-learn.
Try it out!
