Aditya Vyas

New Brunswick, NJ

A Data Science enthusiast interested in applying statistical and predictive techniques to the stockmarket.

   About Me    Skills    Experience    Projects    Contact Me

Data Scientist

Working at the intersection of Machine Learning and Finance

Developing trading agents using Reinforcement Learning

Using combination of R and Python for data analysis

Do you think I am good for you? - Resume

PROGRAMMING LANGUAGES I KNOW
MACHINE LEARNING TOOLS AND LIBRARIES I USE
OTHER TOOLS AND LIBRARIES I USE
Deep Q-Learning Agent for Stockmarket Trading
Predict Stock Returns using GLoVE Embeddings and Document Vectors
Student Educational Data Analysis using Spark
Table Header Detection and Classification using Machine Learning
Recurring Record Extraction from Document Images using Fuzzy Logic
Mining of Professional Details using homepage Web-URLs
Event Management Web Application using Flask
Trust Region Policy Optimization (In Progress)
Multi-Agent Synchronous Advantage Actor-Critic
Single-Agent Synchronous Advantage Actor-Critic (A2C)
Asynchronous Advantage Actor-Critic (A3C)
Deep Deterministic Policy Gradients (DDPG)
Deep Q-Networks (DQN)
REINFORCE with Baseline
SARSA with Linear Function Approximator
Q-Learning with Linear Function Approximator
Tabular Q-Learning(λ)
Tabular SARSA(λ)
Double Q-Learning
Q-Learning Off-Policy TD Control
SARSA On-Policy TD Control
TD(λ)
TD(0)
Monte Carlo Control with Exploring Starts
On-Policy Monte-Carlo Control
Monte-Carlo State Value Approximation
Off-Policy Monte-Carlo Control with Importance Sampling
Off-Policy Monte-Carlo Prediction with Importance Sampling
Quora Duplicate Question Pairs Competition
Mercedes-Benz Greener Manufacturing Competition
Santander Value Prediction Challenge
Costa Rican Household Poverty Level Prediction
EDA of US Airline Twitter Data
EDA of Beer Recipes Dataset
EDA of Board Games Dataset
EDA of Global Terrorist Dataset
EDA of Spooky Author Dataset
  REACH ME AT
  Email-1
  Email-2
  Skype
: aditya1702
  PLACE I LIVE

Apteo                                                               August'17 - August'18

Worked as a data scientist at this fintech company based in Manhattan, New York. My tasks included:

  • Developing trading agents using deep reinforcement learning for deciding optimal trading strategies. I created a Deep Q-Network algorithm for executing trades in Apteo’s stock market environment to learn buy, hold and sell strategies.
  • Use financial text articles to generate document vectors (doc2vec) and improve neural network accuracy in predicting stock returns.
  • Feature integration into Apteo's AI investing tool Milton. These features are a combination of company fundamentals data and features extracted from financial articles.

Teaching Assistant (DAIICT)                       January'18 - April'18

Worked as a Teaching Assistant for Prof. Jaideep Mulherkar in the course Computational Finance (CS401) - a course based on the computational aspects and mathematical aspects of stockmarkets. I developed and graded 60 Computational Science students’ weekly python lab assignments related to:

  • Binomial Asset Pricing
  • Risk Free Arbitrage
  • Monte Carlo Theorem
  • Black Scholes Equation

Playpower Labs                                              Sept'17 - November'17

I was a data science research intern and dealt with student educational data. My task was to analyse educational data and gain insights into factors which affect the scholastic performance of students - fluency, speech, endurance and teaching methodology in the schools.

Shipmnts                                                          May'17 - July'17

Worked as a machine learning intern at Shipmnts, an Ahmedabad based logistics startup. Being a logistics company, Shipmnts recieves large number of shipping bills, invoices and import/export documents. My projects were -

  • Image and PDF enhancement using OpenCV for efficient image processing. This involved automated image/pdf degree rotation, removing smudges and producing a high-resolution document as an output.
  • Created a novel algorithm to detect repeated structures in an image from a single annotated instance of the record. The information is extracted with 90% accuracy using Fuzzy similarities, visual and semantic heuristics.
  • Detected table headers in images and documents using machine learning. I constructed features and models to detect table headers in images and pdf documents. The final random forest model achieved a 97% recall by using techniques such as undersampling, oversampling, and SMOTE analysis.

Machine Learning Competitions                  November'16 - Present

I started my machine learning journey at this time by participating in Kaggle competitions and Analytics Vidhya. My entire december was spent working on interesting datasets, interacting with other data scientists on the discussion forums and reading about state-of-the-art techniques of Kaggle grandmasters

Google Code-In                                               November'16 - January'17

Served as a mentor for 30 pre-university students and guided them in tasks involving bug fixing, utility feature incorporation and test case enumeration. The tasks involved helping students write code in CSS, HTML, NodeJS, python's web framework Flask, dealing with Docker images and REST APIs.

Google Summer of Code (GSoC)                   April'16 - August'16

I was selected in Google Summer of Code to work with FOSSASIA as a python web developer to rebuild it's server project open-event-orga-server from scratch. The project was built using python's web framework Flask and this was my first real development experience.

FOASSASIA                                                       November'15 - April'16

Started open-source contributions in FOSSASIA's open-event-orga-server: an event management application for managing events in real time. Solved bugs, raised issues and improved documentation.