Gimlets: Usable Machine Learning
The project aims at making the application of machine learning
algorithms to sensor data more usable and human friendly. My specific
interest in the project is to understand and apply how interactive
visualisations help the goal of usable machine learning.
Machine learning classification of sensor data is powerful- it lets
you build systems to do activity recognition, stress detection,
emotion recognition, and a lot of other tasks (especially
physiological ones). But it's hard for non-experts to build
classifiers. In the first part, we built and user-tested Gimlets to understand why
it's difficult and what features could help future systems.
am investigating the role that visual analytics can play to help
understand and analyse the results of a classifier. This will make the
relationship between data, algorithms and the results easy to
interpret. I am working on this project as a part of my research
internship at the Human Computer Interaction Institute in CMU,
May 2013 - Present | Mentor: Anind K. Dey
Breakdown Latency Study
This work measures the impact of system latency on user task
perfromance and cognitive load.
More to come soon!
Nov 2013 - Present | Mentor: Brandon, Anind K. Dey
Edit distance Modulo Bisimulation: A Quantitative Measure to Study Evolution of User Models
In order to analyze how users learn to use new devices, our proposed
metric allows for a progressive comparison of evolving user models
towards the device target model by measuring the behavioral proximity
When a user learns to use a new device, her understanding of it
evolves. A progressive comparison of the evolving user models towards
the device target model, for analysing learning, involves determining
the behavioral proximity between them. To quantify the gap between a
user model and a target model, we introduce an edit distance metric
for measuring their behavioral proximity using a bisimulation-based
equivalence relation. We define edit distance to be the minimum number
of edges and states with incident edges required to be deleted from
and/or added to a user model to make it bisimilar to the target
model. We propose an algorithm to compute edit distance between two
models and employ the heuristic procedure on experimental data for
computing edit distance between target and user models. The data is
organised into two experiments depending on the device the user
interacted with: (a) a simple device resembling a vending machine and
(b) a close to real-world vehicle transmission model. The results
validate our proposed metric as edit distance converges with
progressive user learning, increases for erroneous learning, and
remains unchanged indicating no learning.
Please find the paper here.
May 2012 - Dec 2013 | Mentor: Venkatesh Choppella, Anind K. Dey |
In the Proceedings of CHI 2014.
Mobile Suitability Heatmap
Developed and documented Mobile Interface Guidelines; built a mobile
application for allowing CA Inc. products to check their mobile
The increasing penetration of mobile internet has made it important
for every service provider to have its presence into the mobile
application space. The project aimed to develop a set of mobile
interface guidelines to suit the mobile needs for CA mobile
applications. Accordingly, we studied the factors that drive the
standards for mobile interfaces and issued a set of the required
The screenshots of the mobile application: (a) Home page
of the application. (b) Menu screen displaying the different parameters
used to predict the mobile suitability of the application.
(c) The subparameters for one of the specific parameters in detail.
© 2013 CA Inc.
Further, we went ahead and built an application to verify if a mobile
application is suitable to be used on a mobile device. The application
allows a developer to specify which of the mobile interface guidelines
(suggested by us) are followed by his application. Each guideline is
further divided into multiple parameters to check the extent to which
his application follows that particular guideline. Our application
then generates a heatmap to suggest how much the developer’s mobile
application is suitable for the mobile platform.
Jul 2012 - Mar 2013 | Mentor: Kiran, Rama Velpuri | CA Inc.
Functionality or User Interface: which is easier to learn when changed?
A study to compare the learning gaps that are introduced when a user
is exposed to a new version of a software, with which he is previously
The rapid release model of software introduces frequent updates to the
existing software every twelve-eighteen weeks, forcing a user to get
accustomed to its new features. We propose an experimental study to
compare the learning gaps that are introduced when a user is exposed
to a new version of a software, with which he is previously
acquainted. In order to explore the problem, we propose four models of
a machine, with each model involving an update either to the
functionality, or to the user interface, or both. We conducted a
between-subjects experimental study with thirty-two participants who
performed two tasks successively on two models of a machine, the
second one being a updated model of the first. The analysis of the
data using ANOVA implies that a change in the user interface dominates
a change in the functionality. Results indicate that 88% of the errors
were caused due to a change in the user interface. 87.5% of the users
who underwent a change in the user interface hold this change
responsible for the learning gap, while only 56.25% users who
underwent a change in the functionality consider it to be a potential
reason for the learning gap.
Please find the paper here.
Nov 2011 - Nov 2012 | Mentor: Venkatesh Choppella, Harini Sampath | Published in IHCI 2012.
Virtual Labs User Interfaces
Our team built a framework for designing interfaces and structuring content to
support asynchronous Virtual Lab development.
The project involved working out a solution that allows multiple lab
developers to build user interfaces for their respective labs
independently, such that:
The solution is articulated in detail by our team leader in a publication
- There is no mutual interference from different lab developers.
- They need not worry about any changes made to the main UI managed by the VLEAD team.
- The UIs for all the labs must always follow the standard UI guidelines as suggested by the VLEAD team.
VLEAD (Virtual Labs Engineering and Architecture Division), based in
IIIT-Hyderabad campus, is one of the several teams working on the
'Virtual Labs' project. Know more
Aug 2011 - May 2012 | Mentor: Rohit Ashok Khot, Venkatesh Choppella | Team: VLEAD
Y2Y: Youth to Youth
Designed and developed a framework to collect and disseminate
information pertaining to urban jobs among the rural population
in their local language through SMSes to harness the readily
accessible mobile platform.
Access to digital information is readily available in the urban
sector in India. It is however still a privilege in the rural
India mainly due to the low penetration of computers and
internet. Limited knowledge of English language further restricts
this access. Employment being one of the dearest needs of this
section, we decided to start from information pertaining to jobs
in the urban regions. Based on a few surveys undertaken in rural
areas neighbouring Hyderabad (India), we identified an
information gap between the youth of the rural sector and the
urban areas, that was deepening further. Though equipped with
essential qualifications, these villagers had a little know-how
of the existing opportunities. Apart from the limited access to
web through internet cafes, another major concern was the lack of
access to the right information that the local audiences could
benefit from. As expected, medium of instruction mostly being
English further contributed to this information gap.
The solution was to build a Youth-to-Youth information transfer
system, so that the information available to the urban youth could be
accessed by the unprivileged rural youth. We aimed to bridge the above
identified information gap by integrating language and speech
technology tools developed at various centres at IIIT-Hyderabad. This
integration would result in client-server mobile information system
where the users would get information in an identified domain (like
education, jobs etc.) in their local languages on mobile phones
through SMSes. Thereafter, they may look the references online. My
direct contribution to the project was organising the information
obtained from different sources and in different languages into a
database. I was also responsible for ensuring that each query to the
database is addressed correctly in the minimal possible time so that
the appropriate information could be displayed online or sent via
messages as required.
A rough map illustrating the architecture of our designed solution.
The tentative chosen domain of ‘Jobs and related work opportunities’
can be later extended as a complete information dissemination system
informing about government welfare schemes, health care solutions,
etc. to the rural population.
Jan 2012 - May 2012 | Mentor: Mrs. Kavita Vemuri | Team: Jay, Manushree, Vishal
Demography based Automated Teller Machines
Identified and differentiated the design requirements of an Indian ATM
from the global standards. We also suggested a simple design solution to
help address the identified issues.
Automated Teller Machine (ATM) is a widely deployed and used system to
cater the banking needs of the user. The idea behind availing an ATM
based service was to improve the efficiency of overall financial
transaction process. However, in practice, many users with native
languages other than English, struggle to efficiently utilize the ATM
services on offer. The reason being, even after offering services for
more than 40 years, the design of the ATM remained unchanged despite
the commonly observed shortcomings. Most of the proposed attempts to
redesign the ATM interactivity with the user do not befit a particular
diverse demographic setting. The aim of this paper is to address the
issue of ATM usability in India. We came up with a novel ATM design
both in terms of hardware and software that offers the user a
personalised space to interact with the intuitive interface. The
proposed design when tested, measured a usability score of 79 on the
System Usability Scale, and hence suits the Indian scenario well. The
study also issues a set of guidelines that should be considered for
designing a system specific to a demography.
Here's the video that we presented at USID 2011 held at Auroville.
Video presented at USID 2011.
Please find the paper here.
Aug 2011 - Mar 2012 | Mentor: Anind K. Dey | Team: Deepti Aggarwal | Published in IHCI 2012. | Design accepted in USID 2011.
Bingo: A smart trash bin
A working prototype of a sensor operated trash bin that enlarges its
effective input surface area upon detection of any oncoming waste.
Damn! There again the trash lay off the container! Often many of us
are crestfallen at our unsuccessful attempts to throw garbage into the
trash cans. Not to forget the umpteen times when you are too lazy to
move and take care of the trash lying around. Also, we were not too
keen to see the trash can creep around. This was mainly since, mostly,
our thrown trash ends up quite close to the trash can, but inside
it. We blame this failure upon the limited opening of the trash and
propose "BinGO" to serve us lazy folks better!
We wanted something that blossoms and shrinks alike a flower. The
idea further evolved to having an inverted umbrella atop of the trash
to help enlarge the effective opening. But how does this umbrella
unfold itself in a corner of a room? Considering the limited space
around a trash can, following images showcase how our idea evolved to
a square form.
The proposed form with a square top initially opened from the
center. This obstructed the trash that the can intended to capture.
It was here that the concept again underwent a serious of design
changes till we came up with the final concept.
We completed the design with sensor operated flaps that opened up to
enlarge the effective opening upon detection of onwards coming trash.
Considering the time constraints of the workshop, we limited ourselves
to two flaps- the front and the back. The prototype was built using
two servo motors that were operated using sensors. Our prototype
worked effectively for trash thrown from a distance of about 3 feet at
the same level of the opening.
We presented this project in MIT Media Lab workshop, Delhi, India in
2012 in the sesor-mediated track. (Apologies for the missing video!)
Mar 2012 | MIT Media Labs Workshop, Delhi
Mobile Interface to Support Users of Varying Skill
Prototyped a mobile interface that supports both novice and expert
users to allow sharing of mobile phones within the economically poor
sections of Indian society.
The project involved choosing a device used regularly by people who
find it difficult to operate, and improve it to suit their needs. I
often saw senior citizens of India facing problems in using mobile
phones mainly due to the technical inadequacy. Their lack of trust in
technology further discourage them from using it. Moreover, often
mobile phones are shared by people of a family in poorer sections of
the society. This neccisitates a mobile interface that supports the
interaction practices of novice users without making it frustrating
for the expert users.
The iterative design process mainly involved the following actions in
the specified sequence.
First Design Iteration: A think aloud session helped me to uncover
the general issues faced by people belonging to both the groups of
mobile phone users- naive and expert. Based on the information
gathered in the think aloud session, we presented a paper based
prototype for the interface.
Second Design Iteration: I gathered user feedback on the first
design prototype that we presented to the users. Based on this, we
modified the design and built a software prototype for the
Third Design Iteration: I performed a heuristic evaluation on the
designed prototype to have a detailed insight about the usability
issues within the design formulated in second iteration. The
realisations from heuristic evaluation were incorporated into the
Fourth Design Iteration: I conducted user testing on the prototype
built in thrid iteration using a specified set of tasks and
improved the design as per our findings of the user testing.
Fifth Design Iteration: A final round of user testing was conducted
to take care of any issues or limitations left in the design from
Aug 2011 - Dec 2011 | Mentors: Anind K. Dey, Jen Mankoff
Designing effective online examinations: Guidelines from a HCI evaluation
Assessed the online examination system of software engineering course
offered at IIIT-Hyderabad; issued a set of guidelines from a HCI
evaluation and incorporated them into the existing system.
The various set of arguments in favor and in opposition to the
introduction of web based examinations have made it essential to check
the skills required to master this transition. It is this ideology
that has made comparison between the earlier (paper based) and current
day (web based) methodologies significant. Therefore, we conducted a
study on the online examination portal used by the students of the
software engineering elective offered IIIT Hyderabad. It consisted of
a survey that gave a basic idea about the student perception of the
portal, followed by interview sessions with the students to
acknowledge their issues and concerns in certain specific domains with
respect to the usability of the portal, and later an experiment
supported by talk aloud assessment. We conclude the report by
discussing the usability techniques that will help us formulate a set
of guidelines to be followed and thus design better online
Please find the report here.
Aug 2011 - Nov 2011 | Mentor: Herre Van Oostendorp, Bipin Indurkhya | Team: Kirti Garg, Abhinash
Principles of Programming Languages Virtual Lab
Designed and developed a Virtual Lab for Principles of programming
The POPL virtual lab is a part of an MHRD iniitiative to build virtual
labs to provide remote access to labs in computer science, To enthuse
students to conduct experiments by the arousing their curiosity, and
to share costly equipment and resources, which are otherwise available
to limited number of users due to constraints on time and geographical
You can find more about the project and an online version of the
finished lab here.
Aug 2010 - Apr 2011 | Mentor: Venkatesh Choppella | Team: Swathy Pendyala, VLEAD
Designed and developed a prototype for a concept idea involving open field version
of an indoor game 'Carrom'.
Please find the
Jan 2011 - April 2011 | Mentor: Kavita Vemuri | Team: Vishal, Shashank, Jaspal
Hostel occupancy map generation
A user friendly, web-based portal for better administration of the
hostels; included features like room booking, automatic batch
allocation, key handling and complaint registration.
This was a course project for the Software Systems Analysis and Design
coursework. The aim was to smoothen the process of room allocation
given the limited availability of hostels on campus in IIIT-Hyderabad,
India. Apart from centralised room allocation, we also took care of
registration of complaints and archiving the maintainence
activites. The process involved requirements gathering, design and
development of the software, and verification and validation of the
built software. We took support of different tools and languages
Mar 2010 - Jul 2010 | Mentor: C.V. Jawahar, Kirti Garg | Team: Vishal Garg
Ekjaa website layout
Designed the information layout and the web pages for Ekjaa organisation.
Also designed the logo and various posters to help the Ekjaa team
Please visit Ekjaa here!
May 2010 - Jul 2010 | Team: Harsha Mukherjee, Tushar | Ekjaa.org
Developed a plugin to automate the process of importing the reST wiki
pages into trac database.
May 2010 - Jul 2010 | Mentor: Venkatesh Choppella | Team: Shilpa Sarawagi
Teaching Assistant Portal
An online portal to rank teaching assistants as per their applications
to streamline and automate the process of selecting a few most
suitable assistants for a particular course from the pool of
Jan 2010 - Mar 2010 | Mentor: Kirti Garg | Team: Ishaan Singh
An application that uses Yahoo map APIs to allow tagging a place (like
Hospitals, Schools, Restaurants, etc.) with its features to optimize
the search for such facilities.
Our work at Yahoo Hackathon organised at IIIT-Hyderabad, India, in
2010 secured a place within top 10 among 150 teams. Often, we want to
visit a restaurant not because it is in the vicinity, but because it
serves a particular cuisine. The same extends to different facilities
like hospitals, shopping places etc. which can be ranked as per
different features available at that facility. Our hack focussed on
allowing the users to tag a place with its features and then optimize
the search for a facility based on its specific features.
Jan 2010 | Team: Mohak, Navni | Yahoo HackU
Research Visitor at Ubicomp Lab, HCII
May 2013 - April 2014 | Carnegie Mellon University | Pittsburgh, USA.
Intern for Innovation Center CA Inc.
Jul 2012 - Mar 2013 | CA Technologies Inc. | Hyderabad, India.
Research and Academic:
Research Assistant for VLEAD
Aug 2011 - May 2012 | VLEAD | Hyderabad, India.
Teaching Assistant for Principles of Programming Languages
Aug 2011 - Dec 2011 | IIIT-H | Hyderabad, India.
Research Assistant for Virtual Labs
Aug 2010 - Apr 2011 | Pascal Lab @ IIIT-H | Hyderabad, India.
Teaching Assistant for Humanities
Aug 2010 - Dec 2010 | IIIT-H | Hyderabad, India.
Designed and Developed web pages for The Development Stories (project discarded)
May 2012 - Jul 2012 | Hyderabad, India.
Designed web pages for Ekjaa
May 2010 - Jul 2010 | Hyderabad, India.
Times when I am free
I am a keen traveler, but need to visit many more places before I mark them on a map! :(
Making cards, art, and origami is another one!
I also like studying spatial design and huge architectural spaces.
Buildings surely drive me crazy! Apart from appreciating,
I do sketch decent floorplans, elevations, interiors etc.
and end up doing it most of my spare time.
Food, something I can never get bored of! Also try cooking at times.
That's all about me for now! :)