Making your YouTube experience awesome-er !

Worked on building statistical models & deep learning model architectures:* Constructed a CNN+LSTM architecture using CTC loss (on variable word length) for name field Handwriting Recognition.* Coded a CNN model for multi-digit Handwriting Recognition using Goodfellow loss that achieved 96% word accuracy.* Implemented Face Recognition (used MTCNN for face extraction, Siamese network & K-nearest neighbors for embedding comparison) from Yearbooks collection for Mobile team and optimized for low latency inference (< 4 secs)* Built the end-to-end ML platform for deploying all our business models [customer segmentation & churn models (US & UK)] into production, with detailed metrics logging & custom Slack notifications in case of warnings & errors.* Built the Life Time Revenue model using Survival Analysis, programmed in Python: It gives a forecast of the total dollars that will be generated as revenue in next N years from each of our active customer. It enables the Finance team to take appropriate business decisions well ahead of time, providing an estimate of dollars we can bid on each customer. The model output distinguishes high valued users from the rest, and the forecast per user can even be used as an important feature in other business ML models to retain customers.

Worked extensively on feature extraction for machine learning models & productionizing them in a cost efficient manner using various AWServices.

Worked on building scalable machine learning applications, and solutions for distributed deployment of models on production:* Developed a service (Java Spring Boot) which takes obituaries and extracts all names, gender, places, dates, along with their corresponding relations, using an NLP model. Deployed on Kubernetes cluster and ran 50M obituaries on production* Introduced TensorFlow Serving on GPU and transformed the deep learning model deployment pipeline, thereby reducing deployment time by 50% along with space complexity, and enhanced overall efficiency by over 80%* Developed a custom monitoring tool using New Relic API to troubleshoot high-throughput deep learning models running on production* Devised a pipeline for converting a model from one framework to another using its computation graph & trained parameters* Created an automated solution to crawl competitor's prices & report them to the marketing team

Planned and developed online sweepstakes and campaigns using PHP & MySQL on the Laravel framework as backend languages, and HTML, Javascript, jQuery, AngularJS, CSS, SASS as the frontend languages. Provided extensive insights on site availability, views, shares & other reports using Google Analytics. Extensively used Facebook, Twitter & Instagram APIs to increase users participation. Experience with AWS, REST, Mechanical Turk, Young America, SOAP and Epsilon. Built interactive games and hosted online. Worked on building an e-commerce site for a company product. Ideated and built the Silver Screen for client BIBA. Developed and managed the official company website.

Initial Learning Program in Business Intelligence & Performance Management that included advanced SQL, PL/SQL, Java, Informatica & SAP Business Objects.Worked in projects of Ally Inc. (IBM Datastage & Unix), Ocwen Mortgage Corp. (ASP.net), J&J (SAP APO & SAP IBP)

Had an experience of one and half weeks in each of the three Overhaul shops of Air India, viz, Electrical, Radio and Instrumentation O/H shops. Learnt about each components of each shop in details and helped the engineers and technicians in their work. Also gained knowledge on flight profile and aerodynamics.

Worked on various Autonomous Robotics problem statements using AVR programming on ATMEL 16-bit microcontrollers, taking in inputs as indicated by the IR sensor pairs from one port, performing the necessary actions by delivering respective outputs to another port which in turn gives input to the motor driving IC L293D. Implemented various solutions using Line Follower, Wall Follower, Maze solver, Bot-to-Bot communication and Obstacle Detector. Also worked upon digital sensors.

Completed a full course on Industrial Automation (PLC) and also gained a practical experience of every course content. Implemented the same to build a project on Sequential Switching of Motors.The course comprised of the following subjects:1. Hardwired Logic Control & Relay Logic Application2. Programming PLC by Simatic Manager Language (LAD, STL, FBD)3. Programmable Logic Controller4. Drives (AC/DC), Profibus and Remote Interfaces