Work Experience

MS in Computer Science - IIIT Hyderabad (July 2018)

1) Multi Agent Reinforcement Learning in Pommerman

Pommerman is a multiagent game where you use 2 cooperating agents to play against 2 competiting agents and partial observations of the agents and communication between cooperating agents. We first use the agent provided by them which used graph algorithms to take an action and make an imitation model for our agents. The infrastructure uses CNN to encode the grid data and lstm for the inclusion of previous timesteps and runs a actor critic model to train. Then we play our agents against different kind of agents and our own agents(curriculum learning) to improve our actor critic model .

2) Constructing an AI to reduce errors in track detection in flights PAN India

We first use hybrid clustering and use it along with the input data. We construct a POMDP ( Partially observable markov decision process) structure and learn the transitions through ensemble learning of SVM, xgboost and LSTM which gives the confidence in our data point and then cluster the seperated tracks of a single flight together.

3) Find objects in images

We use a combination of 4 cues - edge density, multi scale saliency, superpixel and color contrast to find the windows containing objects and then use non max suppression sampling to find the optimal windows and train a naive bayes model with a training set of window corner locations.

4) Scene Text Detection

We use stroke width transform in an edge map and use connected components to detect text in natural images like sign boards on roads. We then identify the characters using spacing and recognize words in the text using Resnet neural network architecture.

Data Scientist- LatentView Analytics( Sep 2015 – Jun 2017), Chennai

LatentView Analytics is an end-to-end marketing solutions provider that brings together the power of Data, Analytics, Technology, Campaign management and Digital platforms to help businesses acquire customers intelligently, retain them optimally and manage them profitably.

Project - Customer Retention Management:

a) Structure LVM (CRM) definitions, provide insights for planning campaigns, make design choices for individual campaigns, sample size estimations (Test/Control, Test/Universal Control) using R, UAT, campaign performance tracking using excel, VBA and tableau.

b) Build high value merchant model using libraries like sklearn, pandas, numpy, xgboost in Python, campaign performance tracking using excel, VBA and several tableau dashboards.

c) Independently handle delivery of analytics assignments, co-ordinate with onsite LatentView consultants and client teams to ensure high quality, on-time delivery.

d) Analyzing the impact of the campaigns and come out with data based insights which can help in driving sales and increase customer lifetime value.
e ) Identify new business opportunities with the client, preparing sales presentations and proposals for prospective assignments.

Project – Recommender System:
a) Built a Recommender system for a prospective food delivery client in pyspark.
b) Used libraries in python like flask and cherrypy to create a web service. Used JavaScript to create front end.

Project – DPL Data Team:

Created a web crawler to fetch all of google play data for organizing a competition for a well-recognized analytics competition using scrapy framework in python.

Project – Prospect Conversion :

a) Created a tool in excel using VBA which arranges the prospects and old customers based on their chances to take a loan for a big automobile client.

b) Used Logistic Regression of previous prospects to find the top prospects to call for the sales team.

Trainee - ITW biss (Jul 2014 – Dec 2014),Peenya Industrial Area, Peenya Industrial Area

Succesfully implemented a project on Comprehensive System Validation which reduced the validation time of the most commonly sold testing machine manufactured in Biss by 20% by implementing the skills i learnt from six sigma green belt certification.

a) Business Development - Pitching to companies and startups in Bangalore and Hyderabad which can provide part time job opportunities and virtual internships to students from various colleges from bangalore, hyderabad, pune, mumbai.

b) HR - Getting in touch with students from top colleges in bangalore and hyderabad and assigning campus ambassadors in each of the colleges.

c) Digital Marketing - Publicizing our company website through our facebook page, creating posters , etc.

Marketing and Video Editing Intern - Beehives(2014) Gachchibowli, Hyderabad

a) Market research in Sports, Startups , Social and Film sector for getting customers who may get help from crowdfunding by cold calling, social networking or references.

b) Researching about tools which can help in maintaining the company's database like ZOHO , Microsoft Project Professional.

c) Shooting and editing the videos of the customers who wanted to crowdfund through us.

Intern - WINDALS AUTO PVT.LTD. (May 2013 – Jul 2013) ,Nagpur Area, India

a) Made a fixture for complete 360 degree free rotaion of the workpiece in all the 3 coordinates .

b) Did research on Weld Penetrations samples and gave a presentation of the same.