Announcing the Winners! AI for Education Global HackWeek, 2017

5 winners from 3 continents taking home $19K in prizes + lots of honorable mentions

Thank you to the 1,400 students and professionals from around the world that signed up to participate, to the 350 of you that attended’s free ML and Data Science workshops in-person two weeks ago, and to our partners IBM, Google Developer Groups, Omidyar Network, Digital Ocean, Amazon Web Services, IIT Delhi, Camp K12, and NITI Aayog.

We received projects from 80 participants, and our judges Prof. Alexander Rush (CS prof @ Harvard, head of Harvard NLP), Denny Britz (Stanford, Google Brain,, François Chollet (Google ML research, author of, Joel Tetreault (NLP expert from Grammary, ex Yahoo! Labs) had the very difficult task of narrowing down to the top 5 winners that will take home $19K in prizes.

After much deliberation, the judges have spoken and the verdict is out. A huge congrats to our top 5 prize winners and to our honorable mentions (also included below) who submitted brilliant projects. We unfortunately did not have enough prizes available to reward all these great projects, but we want to give a shout out to all participants who dedicated their time to working on AI for Education over the past two weeks and open sourced their work.

$6,000 Grand Prize (sponsored by ON)

Winner: ReadEx
Mayank Rajoria, Prakhar Gupta, Harsh Arya, Prakhar Agrawal from IIT Delhi
Description: Android app that does real-time question generation using NLP, content recommendations, and flashcard creation (for questions you answer incorrectly) as you read to create an engaging and adaptive learning experience. Search for content on the web, Open PDFs / files on your phone, or scan images using OCR to load text and start reading.

$3,000 Second Place (sponsored by IBM)

Winner: GradeGuardian
Andrew Arpasi, Andrew Rothwell, and Vishnu Vijayan from Purdue University
Website / Live Demo:
Github org with multiple repos:
Description: Predictive models and visualizations for student performance with an interactive dashboard showing anticipated effect of policy changes. Submission includes 3 components packaged as a single web app — a Chatbot that inputs student information, an Advisor Console that shows students at risk, and a prediction module for government / policymakers.

Live demo at

$5,000 AWS credits, Third Place

Winner: StyLing
Anusha Ravi from Oxford University
Website / Live Demo:
Description: Learn how to write fluently in Hindi and other Indian languages. This demo features a Hindi language model built from scratch using Keras/Tensorflow with Facebook’s fastText Hindi word vectors and a Hindi language corpus of 50k words. The app compares the perplexity of the language model’s distribution on the provided user input to the perplexity achieved on the original corpus as a measure of fluency. The predicted probabilities for each word in the sequence are used to create a heat map highlighting problematic areas.

Try out StyLing at

$3,000 Google credits, Fourth Place

Winner: SAT Buddy
Chirag Mahapatra (UC Berkeley), Karthika Purushothaman (CMU), Devashish Sharma
Live Demo (FB Messenger Bot):
Github Links:
Description: Conversational interface for learning Math, built on the FB Messenger platform. Capable of: delivering tests with multiple choice questions to students, recommending relevant youtube videos in respond to a student’s input query (via a custom ML model trained on Keras), solving equations step by step (Wolfram Alpha API), handling misspelled queries from users (Bing API), and showing students personalized learning analytics (server-side image rendering), and having general conversation (

Architecture diagram for SATBuddy

$2,000 Google credits, Fifth Place

Winner: Text to Narrated Film
Gopi Swamy Veerandra from IIT Delhi
Generate videos from PDF/text through topic identification, image search, and text-to-speech. Text is boring — this system auto generates engaging narrated films from PDF textbook excerpts as follows:

  1. PDF to text conversion, and text cleaning / pre-processing
  2. Topic identification and extraction using NLP (could use LDA modelling in the future)
  3. Search images by topic and download from Google Images
  4. Create Audio for sentence using Google text to speech library
  5. Create MP4 video from images + text + audio

Video of sample output (auto-generated from this pdf):

Honorable Mentions

1. Most impressive custom NLP model

Project: Named-Entity Recognition system in Hindi using Conditional Random Fields
Team: Sushant Rathi & Shivam Shaswat
Video: see below

2. Creative application of Computer Vision to Education

Project: Automatic attendance marking using OpenCV
Team: Deepanshu Geed, Atul Daluka, Puneet Mittal

Footage of students emerging from a classroom auto-tagged with name of student

3. Best use of pre-trained ML model

Project: Instant question answering for currently viewed text content and MOOCs using an open source machine comprehension model
Team: Saksham Soni, Prashant Mittal, Krittam Kothari

Chatbot answers questions that students have while reading and provides much needed interactivity

4. Best UI/UX

Project: 4Cards, Hindi vocab learning React Native app
Team: Pratik Soni & Ranjith Kumar

5. Youngest AI for Education hacker

Project: AI enabled Learning Lab featuring Alexa skills and conversational agents created with Tensorflow
Team: Sayli Bande (9th grader!)

6. Best use of IBM Watson API

Project: Generate relevant news articles for content the student is currently viewing using IBM Watson NLU API to extract key concepts from text and using IBM News API for retrieving related news clips.
Team: Nirant Kasliwal & Ankit Bansal
Live Demo / Website:

7. Best Data Science visualization

Project: State-wise and district wise analysis of education data to determine factors impacting school dropout rate
Team: Aditya Thakur
Code + Report:

Mapping changes in student dropout by district of India, 2012 to 2015. Full report at

8. Best Data Science write up

Project: Two part statistical analysis of education data in India: analysis of short term factors affecting dropout rate in schools and modeling of factors affecting long-term engineering skill
Team: Sunayana Raye (MIT) & family
Code + Report: