Data Scientist with hands-on, managerial, and business  experience. 

I have been working at NYC based boutique data science consultancy, Byteflow Dynamics,  on high impact projects across healthcare, tech,  and finance (over four years experience).  In the  first two years I was more hands-on:  building models; dashboards; machine learning algos; research & development; proof of concept products.. In the more recent 2+years I have been involved  in a combination management & business side: managing small teams, hiring and onboarding, upskilling and training;  identifying new lines of business; on-boarding of new clients and business development. 

 

Formal training: Physics PhD, Quantum Information Theory, bringing some of those techniques to data science & business. Interested in an impactful DS position collaborating with tech and business stakeholders.

WhatsApp%20Image%202019-06-05%20at%2018.

Experience, case studies, and more 

Case Studies: 

Hedge Funds: data scraping; natural language processing; deep learning 

Healthcare: member segmentation 

Insurance 

Experience: 

Entrepreneurship 

Interests 

How Andi's services may help you

Training 

AI Leadership 

New Product Development 

Advisory 

Contact info:

ashehu@byteflows.com

https://www.linkedin.com/in/andi-shehu/

Text: 347-278-2892 

Hedge Funds

 

Sentiment Analysis and data scraping

In the world of finance traders and analysts are always vying to stay ahead of the curve when picking stocks using any data sets that they can get their hands on. This can be in the form of time series which can be purchased from the likes of a Bloomberg terminal or it can be an earning reports. 

In addition, many of the large institutions deploy an army of analysts to analyze news coverage on particular stocks. In the old fashioned world, this is done by manually going through articles covering a list of stocks and counting positive and negative words for the overall sentiment. A final score is calculated using some internal proprietary algorithm. The score is handed over to the traders who then decide whether to buy, sell, or short a partiular stock. 

There are however a few challenges with this methodology:  
  *i* it is not a scalable solution, often firms will limit the number of sources and stocks to follow;    
  *ii* humans make mistakes, they get tired, and auditing an analyst's work adds to the cost of production.    

 
## Problem solution in brief 

Text analytics and Natural Language Processing (NLP) allow for automation of existing scoring methodologies, scaling them up, and quickly iterating on newer models. Once the model is up and running, it can be scaled to score thousands of articles at once. Even though these methods have existed for decades, what is new is the tools from cloud computing and open source community such as R and Python programming languages. We now have the necessary tools to scrape news sources from top blogs and sites. 

Here, we present novel approaches to sentiment analysis over one hundred stocks. The data is scraped from ten news sources over the period of a week and the sentiment is calculated for that week.
 

Deep Learning, Machine Learning 

135 West 41 Street. 5th Floor. New York, NY 10036. USA

©2020 by Byteflow Dynamics