Manjeet Kumar

End-to-end implementation of a central platform for event onboarding, model/report management. Business development, planning, designing, leading, and implementing technical solutions and improvements. Created an ML system to predict a query run time on the engine. It reduced the resource utilisation up to 20%. Designed a data warehouse to store and analyse 100 Petabytes of data. It includes ETL process and data consumption as models. Prism Framework: All in one user facing framework to increase Analyst productivity. It involves Airflow as scheduling tool, REST APIs service and User Interface. Users can onboard events via creating pipelines, create models/reports to aggregate and move data from one layer to another layer. Fail Fast: An ML based algorithm to predict the query run time and fail the query without running on the query engine. It uses 20-30 features to predict the run time. Started with an analytical formula in phase 1 and involved ML models in subsequent phases to achieve the target.

Involved in ETL process, handled around 10-15TBs of data per hour. design, architect, implement, and support key datasets that avail structured and timely access to actionable business insights. Created warehouse on Hive backed by AWS S3, moving from Redshift to Hive and Presto. Data pipeline & Warehouse Management: Tracker-based preparation of knowledge base for various platforms and properties of organisation, where user analytics is aggregated using Spark. A server side application later processes this knowledge base to further implement a personalization engine based on user-preferences. Which leads to enhancement of user experience from video line-up to personalised video feed.

Data warehouse setup and building intelligence with data for news personalization. Worked with NLP tools to organise text data and metadata extraction. Learned HLL++/Minhash buffers and widely popular Bloom Filters in data storage and modelling. ROCQ : Analytics: rocQ is a mobile app analytics and user engagement platform that provides real-time actionable insights to businesses. It helps app owners use data to manage user interaction with their apps and build a relationship with their user base. Auto Tagging Tool: Tagging tool suggests probable candidate tag for text content/story using NLP, Wikipedia and DBpedia data. It also takes care of spelling correction and deduplication of suggested tags with weight associated with each tag based on various parameters.

Understanding the client requirements to formulate a project model and communicate the progress and updates to the client on regular intervals. Take care of the complete project life cycle from its inception till the final deliverable to ensure that requirements are met. Big Data Text Analysis: An analytics reporting tool for an input corpus that includes metrics like: word frequencies, topic and keyword extraction. Further the user can use the tool to perform concordance, prepare word sketches, find entity-relationships, and search for related parts of speech across the corpus for a given text. Demographic Prediction: In this project we predicted the demographics of an online user by looking at the browsing history of the user. Data mining and scraping techniques were used to collect data from the online sources. The model was built on a mix of Support Vector Machine algorithm and Naïve Bayes Algorithm. Social media comment’s Analysis: The essence of this project is to analyse the past trend of sentiment for properties in a Hotel chain. Dependency parsing, POS tagging techniques were extensively used in this project along with English language rules to extract data from the comments. Web Analytics: The core functionality of this project is to give an analysis of a webpage, about its user, device detail, geographical detail, etc. Just integrate a code snippet to the web page and get all the analytics you need. Customers can set notifications for various metrics.

Designed and implemented the backend and frontend from scratch in Scala, Play Framework, and MongoDB for advertisement server based on CPI model and admin UI to manage the campaigns and reporting. Enhanced and maintained existing modules, components and controls. Coordinated with third party development teams, code-merging activities, debugging and fixing of issues. Ad-Server: It stores advertising content used in online marketing and delivers that content onto various digital platforms such as; websites, social media outlets and mobile apps, Based on ad request. Real-Time Reporting System: UI application to see the reporting based on ad serving and content delivery. It shows the real-time report for various matrices. System was built using Kafka, spark as a big data tool and Scala as an application language. Device/Browser Fingerprinting: Algorithm based on user’s device details to assign a unique id to that device as a fingerprint id. System Accuracy was near about 60-65 %.