I am Umang. I am a Data Scientist/Deep learning Engineer at Deloitte Consulting and author of a book on Deep learning. My whole life has been a roller coaster ride so far.
I have worked on a number of problems in my life ranging from Computer vision, Anomaly detection using generative deep learning,Time series forecasting using Deep learning, Reinforcement learning and so on.
I started my research career at the end of my Sophomore year, when I was selected for National Science Academies Summer Research Fellowship Program .I joined an Indian Govt Research institute called IUCAA
I later worked with CERN as a project student on the Higgs Boson Project.
I have also been the official Speaker of Google's Machine Learning Crash Course and got featured in the Google's newsletter few times.
I started web development when I was in 6th grade (11 years old) in 2006 , the website I created is still live ! and can be found HERE This one is a small personal introduction in the words of eleven year old. I created other websites too that year
Later on I joined a Rock band and was the lead keyboard player in it ! , I held junior diploma in music in 2009.In 2012 I started my Engineering Journey .
Worked on a research project involving Scientific Computing, Simulations and High Energy Astrophysics,.The project title was "Monte Carlo Simulation of radiative transfer through magnetic plasma". I used different scientific computing tools such as GNU plot,Mathematica etc and wrote code for simulation right from the scratch in the language C++ I also did algorithm Designing while collaborating with some scientists there as a part of my project. Apart from my project work I also helped a number of Graduate Students with their research problems.
Worked on the Higgs boson Decay Channel H->ZZ->4l. Simulating the Proton Proton collisions of the Large Hadron Collider (LHC) at the centre of mass energy 14 TeV using the event generator Pythia and analysing the data generated through CERN's data analysis framework ROOT. Also considering and generating various major reducible and irreducible backgrounds in the channel for my analysis (ZZ* ,ttbar etc). Wrote the code right from the scratch in the language C++ .
Background process generations are used for analysing and removing the similar processes which are replicating the Higgs Decay Channel .I've been working as a freelance startup generalist for various startups incubated at Delhi NCR and few other places , advising them on various things like requirement elicitation , work flow, web technologies, design and many more things .
Created a people counter with computer vision and Raspberry Pi .I used OpenCV and some other available codes from Github. The people counter uses the technique of blob tracking and MOG( or Mixture of Gaussians) technique to recognize people and their features from the images sent by the camera embedded at raspberry Pi board, the events are sent to a SQL server and plotted using ChartJS Library.
I was a part of the Accenture Digital's Internal Product Development Team which was building Data Science and Analytics applications in Insurance domain .My responsibilities include the initial research, selecting/developing the right machine algorithm, developing Full stack (both UI/UX and back end)of the ML application and also doing the Big data computations. So far I have:
- Built Regression Trees,Ridge regression(and few more ML algorithms) based Predictive analytics applications from scratch using Big Data,Java,Python,BackboneJS,HighCharts JS, UnderscoreJS and PySpark, enabling clients to see both predicitive and descriptive analytics
- Built Naive Bayes algorithm based chatbots using DJango Framework and JS to enable easy interaction with users and enabling them see social media sentiment analysis too.
-Built NLP based Sentimental analysis applications using multiple NLP techniques, Python and Data pulling APIs to classify tweets of a particular twitter handle as positive,negative or neutral.
- Have written Python Scripts to automate the Data Pre processing and Analysis of 5 billion rows of data ,which was 20% faster than previous method.
- Developed automated flows of data science from fetching to cleaning to dumping on BigData to analysis of large amount of data using batching algorithms
- Integrated 3rd Party Data visualization tools to our Analytics ML Applications so as to provide better analytics and descriptions to the users.
- Prepared codes in multiple languages so as to automate a lot of Big data operations in our Big Data Environment which decrease reading operation time by 40%.
- Have written bash scripts to automate the data fetching from multiple sources for fast data collections increasing speed of data collection by 50%
Ready for my new challenge starting April 2019
Authoring a Book on Deep learning! the book will involve a lot of TensorFlow(Not in python ;) and a lot of teaching on end to end application development
The Languages ,Frameworks and technologies that I am well conversed with are: