Printing tool Download PDF

Work experience

Sep 2017Present

Senior Data Scientist - Deep Learning and Computer Vision Lead

Opta Information Intelligence
  • Lead a team of three ML Engineers on building various deep learning models on Machine Vision and NLP
  • Developed Machine Vision and object detection models to predict the floor area, number of garages, number of floors, etc of a house based on Google Street view and Google Satellite View images 
  • Developed a Mask RCNN model to calculate the surface area of a home in a Sattelite image by using the predicted Mask Coordinates. 
  • Saved over $3 Million dollars in labelling costs by building image recognition models for real estate
  • Built Chatbots using Dialog flow for the claims division at Opta, helping make the claims center operate virtually round the clock
  • Build Recurrent Neural Network based time series models to predict the sales to listings ratio for up to 18 cities in Canada
  • Configured cloud infrastructure for Deep Learning, including setting up GPU instances on AWS and Google Cloud with Tensorflow, Keras, Torch, and OpenCV
  • Develop Flask applications to deploy Machine Vision and NLP models into production
July 2018Present

Part Time Founder and Head of Machine Learning

Avastus Analytics (
  • Avastus Analytics is a retail technology startup that uses Computer Vision and AI to create completely cashierless stores
  • Lead a team of 4 ML Engineers to develop Computer Vision algorithms for Retail
  • Set the agenda for all the Machine Learning based development at Avastus Analytics
  • Developed Kalman Filters for tracking a person inside a store using Python
  • Built a Bi-Directional RCNN model for Action Detection in Python and Keras
  • Also developed the Theft Detection, Facial Recognition and Insights visualization using Python and Plotly
  • Raised $120,000 USD in cloud credits from IBM
  • Accepted into 3 accelerators in Toronto
Oct 2016Sep 2017

Data Scientist - Contractor

Bank of Montreal
  • Build Anti Money Laundering models to isolate laundering behaviors using transactional Data
  • Develop clustering and regression-based python scripts to uncover different money laundering scenarios like layering and human trafficking
Jun 2016Oct 2016

Data Scientist - Contractor

TELUS, Toronto, ON
  • Use BASE SAS and Segmented Regression to build interrupted time series models that identify and quantify network upgrades because of Media Optimization
  • Created SAS scripts to do Variable clustering in order to reduce 850 Variables to less than 100 important variables
  • Built classification models in SAS to identify which past customers are most likely to sign up for TELUS business services if offered
  • Created R scripts to analyze Cell tower usage statistics and use dashboards  to share the information to upper management using R Shiny
Nov 2015May 2016

Data Scientist - Contractor

Hershey's Canada, Toronto, ON
  • Extensive use of BASE SAS and R programming to understand the effect of macroeconomic variables such as GDP, Money Supply, Earning Average and some extraneous factors such as trends and seasonality on confectionary Sales
  • Created R scripts to scrape social media data like Facebook and Twitter to analyze and monitor and capitalize on online sentiment towards Hershey products
  • Use Tableau to develop innovative dashboards to convert internal reporting from simple word documents to beautiful interactive visualizations
Apr 2013Sep 2015

Revenue Management and Pricing Optimization Quality Engineer

SAS Institute Inc, Cary, NC
  • Work for the IDEAS division at SAS, working on cutting edge statistical models for the Revenue Management and Hospitality industry
  • Support the development of advanced tools and analytics models using SAS BASE, SAS Macros and develop statistical reports to track, analyze, optimize and forecast revenues, revenue yields for large hotels in the US and around the world
Apr 2010Apr 2013

FM Treasury Analyst III

BB&T Corporation, Winston Salem, NC
  • Supported treasury and other areas of the bank by utilizing SAS and other programming languages and building financial solutions for analysis and decision making. This included tax reporting processes, market risk, VAR, DV01, other risk reporting as needed, etc. 
  • Built SAS forecasting models and conducted time series analysis to analyze trends and calculate credit risk by identifying future direction of swap spreads


Deep Learning

Machine Vision ( CNN, OpenCV, SLAM and Kalman Filters), Sequential Models (LSTM and RNN), Generative Adversarial Networks

Machine Learning

Supervised Models, Unsupervised Models, Multivariate analysis


Pandas, Numpy, Keras, Tensorflow, Caffe, Torch , OpenCV, Sklearn, XGBoost

Databases / Applications

SQL Server, Oracle, AWS Athena, Google BigQuery

Other Languages

R, SAS, C++, Tableau


Sep 2018Present

Self Driving Car Engineer Nano Degree

Udacity Inc. 
  • Detect lane lines from a Car-Mounted Camera using Python and OpenCV.
  • Advanced Lane Detection using OpenCV
  • Behavioural Cloning using Keras and 3 camera positions to drive a simulated car autonomously around a track
  • Build an advanced Multi-Dimensional Kalman Filter for localization and tracking in both Python and C++
Apr 2018Jul 2018

Computer Vision Nano Degree

Udacity Inc.
  • Facial Keypoint Detection System: Built a facial keypoint detection system that takes in an image with faces, and predicts the location of 68  key points on each face
  • Image Captioning Project: Trained a CNN-RNN model to build a complex CNN-RNN image captioning model 
  • Landmark detection and Robot Tracking (SLAM): Implement SLAM for Robot localization and tracking in a 2D world using Kalman Filters
Nov 2017Mar 2018

Deep Learning Nano Degree

Udacity Inc.
  • Bike Share Prediction: Built a neural network from scratch and implemented an MLP, i.e backpropagation and forward propagation using Python
  • Fashion MNIST Dataset: Developed a Keras based CNN to create a simple clothes classifier that uses the Fashion MNIST dataset. 
  • Dog Breed Classifier: Used Keras and Tensorflow based Convolutional Neural Network to implement a multi-class dog breed classifier. Also used OpenCV based Harr Feature classifiers to detect human faces and understand what breed of dogs they look like
  • Generate Simpsons TV Script: Built a recurrent neural network in TensorFlow to process text and then used it to generate completely new episodes of Simpsons, based on old scripts.
  • Generate Human Faces: Built a pair of Multi-Layer Neural Networks and made them compete against each other in order to generate realistic faces, using a Deep Convolutional Generative Adversarial Network implemented in Tensorflow
Aug 2007May 2009

Masters in Management Information Systems

Oklahoma State University, Stillwater, OK

Aug 2003May 2007

Bachelors in Electrical Engineering

Osmania University, Hyderabad, India


May 2008Present

SAS Enterprise Miner 5.2 Predictive modeler 


This certification confirms that the candidate has a firm understanding and mastery of the functionalities for predictive modeling available in SAS Enterprise Miner 7.

Aug 2009Present

IBM Cognos 8 BI Author certification 


This certification confirms that the professional can build reports using relational data models, as well as enhance customize, and manage professional reports using IBM Cognos

May 2009Present

 OSU SAS Data Mining and Marketing Certificate

SAS and Oklahoma State University

The curriculum for this program was designed in partnership with SAS, a leading provider of business analytics and data mining software and services.