Jul 2018 – Jun 2019 · 11 mos

• Built similar video recommendation with spatial features and user interaction
• Item-Item similarity based on user interactions
• Thumbnail based similarity using Transfer Learning
• rank-1 accuracy of 65% and rank-1 accuracy of 90%
• engagement time uptick of 30%
• Designed User on-boarding journey, achieved 10% uptick of retention and time spent
• Diversified videos for cold start.
• Video category affinity for each user is learnt for feed personalisation.
• Built duplicate video detection, precision-0.99, recall-0.90
• key-frame based similarity using Transfer Learning.
• Built watermark detection, precision-0.84, recall-0.70
• Trained CNN for clip and non-clip watermarks in videos.

May 2016 – Jul 2018 · 2 yrs 2 mos · Bengaluru Area, India

• Built Gender Prediction, recall - 0.55, precision - 0.89
• Features were extracted from contact book of users using Indian relationship names dictionary
• Installed Apps, profile photo spatial features were also used.
• Built Location Service using open datasets and geohashing
• Used Opencellid dataset for cell towers and wifi ssids.
• Triangulated lat-long data to generate geohash for unknown cell towers - This helped identifying co-
• located users and POI based user selection.
• Built Language Prediction, this was useful for content localization
• features like user device location, friends location, contact book name language were
• used.
• This model was created for top 8 indic languages on Hike network
• Few features were also generated by running regex on user’s chat on device.
• Built Hike Graph for friend recommendation

Mar 2014 – May 2016 · 2 yrs 2 mos · Bengaluru, Karnataka, India

• Fraud Detection System.
• Deep analysis of fraud cases and user data, coming up with rules and highy scalable learning algorithms
• Built statistical models (NN , Logistic Regression, density based Clustering, N-gram model for monkey typing)
• Set up E2E data pipeline for fraud system, using HBase , Aesop and Redis/Kafka
• Graph :
• Built Flipkart's User Graph.
• Created derived features like strong connected and weakly connected components out of the raw graph.
• Setup Graph data pipeline + Processor engine(batch) for several use cases like Recommendation, Ads, Fraud ( Buyer/Seller)
• Tech Stack for graph( Titan over Hbase, Storm/Kafka, Spark/Graphx)

Jul 2012 – Mar 2014 · 1 yr 8 mos · India

• Worked on creating a distributed grid architecture to process huge volume of data in parallel mode. Used Hadoop and Hive for Data Analytics.
• Created Anomaly detection models to monitor trading activities.

Jul 2011 – Jul 2012 · 1 yr

• worked on creating Cashier System.
• Created eWallet which handles each and every monetary transaction done by gaming or cashier system. It was also exposed as a service to our partners.

Jul 2010 – Jul 2011 · 1 yr

• Senior Software Engineer

Jun 2007 – Apr 2010 · 2 yrs 10 mos

• Worked on various Java/J2ee projects with frameworks like Struts, JSP, servlets, web services, XML