Projects
Deep Social Learning
Individual Social media channels allow us to observe specific behavior of people in realtime. These, taken in isolation, gives us only a part of the picture of what the person is about. Can we integrate the signals a person provides on different channels and paint a complete picture of one’s personna? Even further, can the integration of all the social signals of the Society as a whole result in the emergence of a new type of intelligence.... read more
Robo Brain
RoboBrain is a large-scale computational system that learns from publicly available Internet resources, computer simulations, and real-life robot trials. It accumulates everything robotics into a comprehensive and interconnected knowledge base. Applications include prototyping for robotics research, household robots, and self-driving cars. It currently has researchers working from Cornell, Stanford, UC Berkeley and Brown University. It has received significant press recently. I am primarily responsible for building the large scale databases to represent the brain.... read more
paper Robo Brain: Large-Scale Knowledge Engine for Robots
Ashutosh Saxena, Ashesh Jain, Ozan Sener, Aditya Jami, Dipendra K Misra, Hema S Koppula. International Symposium on Robotics Research (ISRR), 2015 [PDF]
Predict Effect
Predict Effect is a data platform for advertisers to discover relevant audience across social channels for advertising in realtime. PE has analyzed 10's of billions of user social actions across multiple sites (Facebook, twitter, pinterest, reddit, iOS app store etc) for the last 2 years to create a n-layer representation of a user type using deep learning mechanisms. PE has 5 patents (Pending) in the social advertising space and is currently working with few high profile beta clients. I'm co-founder and CTO of the company.... read more
Tell Me Dave
In order for robots to perform tasks in real world, they need to be able to understand our natural language commands. While there is a lot of past research that went into the task of language parsing, they often require the instructions to be spelled out in full detail which makes it difficult to use them in real world situations. Our goal is to enable a robots to even take an ill-specified instruction as generic as “Make a cup of coffee” and be able to figure out how to fill a cup with milk or use one if it already has milk etc. depending on how the environment looks. My main focus has been to scale the associated learning algorithms to millions of data points.... read more
Fan.me
Fan.me is a leading provider of social media analytics to the sports industry, including athletes, agencies, teams and sponsors. Fan has analyzed over a billion conversations related to sports on different social media channels for 2 years produce these sophisticated sports specific interest/topic models for users. Fan's technology has also been used for variety of social causes. Eg, David Ortiz and his marketing team used Fan to raise $200,000 from Social Media users towards OneFund for helping those who were most affected by the tragic Boston Marathon bombings etc. I am currently serving as CTO of the company.... read more
The Netflix Simian Army
Simian Army consists of services (Monkeys) in the cloud for generating various kinds of failures, detecting abnormal conditions, and testing its ability to survive them. The goal is to keep the cloud infrastructure safe, secure, and highly available. It is fully open-source at this point and being used by few tens of startups and other companies running on the cloud. It has got signicant press attention and lot of praises in cloud computing community. I was primarily responsible for building the original framework, Chaos Monkey, Latency Monkey, Conformity Monkey & Janitor Monkey.... read more
PRIAM
Priam is an open source process/tool that runs alongside Apache Cassandra (Highly available column oriented database) to automate the following: Backup and recovery (Complete and incremental), Token management, Seed discovery, Configuration, Support AWS environment. Key features include Token management using SimpleDB, Support multi-region Cassandra deployment in AWS via public IP, Automated security group update in multi-region environment, Backup SSTables from local ephemeral disks to S3, Uses Snappy compression to compress backup data on the fly, Backup throttling.... read more
CassJMeter
Apache JMeter may be used to test performance both on static and dynamic resources (files, Servlets, Perl scripts, Java Objects, Data Bases and Queries, FTP Servers and more). It can be used to simulate a heavy load on a server, network or object to test its strength or to analyze overall performance under different load types. You can use it to make a graphical analysis of performance or to test your server/script/object behavior under heavy concurrent load. This plugin acts as a client of Cassandra and can send requests over either Astayanax or Thrift. The plugin is fully configurable.... read more
Yahoo! Big Data Collection Platform (Data Highway)
Data Highway is a next generation multi tenant hadoop based batch processing system that is responsible for collecting and mining 65 billion events (30TB) of data on daily basis. The metrics generated are used for Business Intelligence, Behavioral targeting of advertisements, accounting (sponsored search ad clicks, display ad views) e.t.c. This is by far the biggest deployment ever done inside Yahoo (and most cloud backed companies) with a hardware foot print of 300,000 nodes.... read more...
Yahoo! Realtime Collection Platform (Data Highway Fastrack)
Data Highway fastrack is a realtime collection platform that is responsible for collecting and analyzing 750K web events every minute from 300,000 front-end serving Yahoo traffic nodes. These metrics fueled the Content Optimzation and Knowledge Engine (COKE) which generates instant recommendations based on user clicks in realtime.... read more
Handwriting Recognition - “Offline” Approach
There are many things we humans have in common. But there are other things that are very unique to every individual - DNA, finger-prints, etc. Handwriting is one other such thing that is unique to every individual, which the recent studies on Handwriting analysis have already proved. Although arguable is this issue, that handwriting can be mimicked and forgery becoming a huge issue, there is certain level of individuality and uniqueness (like the way of holding the pen, the strokes used in the writing and the amount of pressure put on paper, to name a few), that cannot be mimicked or forged. As computerization is becoming more prominent these days, Handwriting Recognition is gaining importance in various fields eg., authentication of signatures in banks, recognizing ZIP codes addresses on letters, forensic evidence, etc.
paper Hnadwriting Recognition - “Offline” Approach
Shankar Rao, J. Aditya. Modern Trends in Intelligent Computing Systems (Nasmotics). February 2005 [pdf]
(Best Paper Award)
Quantum Cryptography
Modern cryptography algorithms are based over the fundamental process of factoring large integers into their primes, which is said to be “INTRACTABLE”. But modern cryptography is vulnerable to both technological progress of computing power and evolution in mathematics to quickly reverse one-way functions such as that of factoring large integers. So the solution is to introduce quantum physics into cryptography, which lead to evaluation of quantum cryptography. Quantum cryptography is one of the emerging topics in the field of computer industry. We focused on exposing some of the weakness of modern digital cryptosystems and implemented a protocol for quantum key distribution. Our work has featured in Cyber Security and Information Systems Information Analysis Center....read more
paper Quantum Cryptography
J. Aditya, P. Shankar Rao. Proceedings of Computer Society of India (CSI). February 2005
(Best Paper Award)