Rajat Bhatnagar

About Me

Hi! My name is Rajat Bhatnagar, I am currently pursuing a Masters in Computer Science at the University of Colorado Boulder. I am currently in the second year. I have completed my Bachelors in Information Technology from Vellore Institute of Technology, Vellore, India. I have 3.5 years of work experience as a Full-stack Software Developer working for a real-estate startup named Fastfox. Apart from that I did an internship at Apple in Summer 2020 as a Software Engineer.

Research Interest

My research interests are broadly in Natural Language Processing, Deep Learning and Machine Learning. I am currently working in the NALA lab under the guidance of Professor Katharina Kann. I am also working with Professor Chenhao Tan in the NLP-CSS Lab at CU Boulder. I wish to research interpretable machine learning and to find out answers to questions like what kind of information is represented in embedding vectors and can we teach a model to understand just a few 100 words properly.

Projects at Apple

  1. I worked on integrating OAuth2 server side authentication in Apple’s internal tool. I designed the data model in mysql and wrote apis for allowing any third party application to authenticate with the internal tool. I integrated this with confluence via script runner which allowed showing the user data from the internal tool on confluence.

  2. I developed the frontend of a utility tool through which administrators could add, remove or update access to system accounts. I used React js and ag-grid framework for the frontend.

Projects at Fastfox

  1. One of the lead members in designing and developing the customer facing website Fastfox.com using React js and Redux.
  2. Developed and designed numerous apis for Fastfox’s internal apps like Broex, Brofirst, LeadGen using Node js, PostgreSQL.
  3. Worked on the recommendation engine for fetching inventories using SOLR. It included factors like user filters, inventory accessibility, inventory age etc. It worked on a mathematical model which gave score to the inventories according to various factors.
  4. Developed a bot to crawl through Whatsapp Web using Python. The bot could perform actions like joining a whatsapp group, retrieve chat messages from a group, retrieve participants of a group, exit a group. All of this was done in a randomized manner such that Whatsapp doesn’t block the bot.
  5. Used the chat messages of real-estate broker from the above task to extract Named Entitites like BHK, price, location, property type (lease / sale), whether the property is a lead or inventory. Used Conditional Random Field for this task to do a Named Entity Recognition. Achieved an F1 score of 81.3%.
  6. Regularly did production releases and was involved in infrastructure management.
  7. Developed Fastfox Payments frontend and backend using integration with RazorPay.
  8. Wrote a crawler in Python for extracting property information from government website.
  9. Integrated AWS CloudWatch with the infrastructure of Fastfox to raise an alarm if CPU, memory, number of requests etc exceeds a threshold.
  10. Wrote various scripts for mocking database, lead flow automation.
  11. Wrote Cypher queries on Neo4J for querying the graph database which was used by Broex.
  12. Worked on sending emails to customers and error logging through RabbitMQ.
  13. Received Fastfox Excellence award in 2017 for performing well.

Projects at CU Boulder

  1. Toxicity Classifier (2020) - Used BERT for classifying toxic comments in Reddit. Developed a website where the moderators can delete the toxic comments. Integrated with the several subreddits. Also highlighted the key tokens due to which the decision was made by the classifier. Youtube Video
  2. Working with Professor Katharina Kann on “New data collection methods for machine translation using monolingual speakers”.
  3. A survey of neural machine translation for low-resource languages. Trained LSTM for 26 languages using low-resource hyperparameters.
  4. Chatting Website (2020) A website for group chat.
  5. Sniper (2019) - Made a system to prevent Denial of Service attacks by blocking the IP address of the attacker. Technologies used - docker, nginx plus, elastic search, elastalert, kibana, filebeats.

Projects at VIT University

  1. Paper publication on MapReduce Based Frequent Itemset Mining Algorithm on Stream Data (2015)
  2. Published Windows Mobile Apps and Windows Desktop Apps (2013 - 2015) - Integration+, My Dictionary, VIT Academia and My Today’s Horoscope

Page design by Ankit Sultana