Hi, I'm Ishaan!

Welcome to my page!

I am an undergraduate student studying Computer Engineering(CPEG) at Hong Kong University of Science and Technology(HKUST). My programming interests primarily lie in Machine Learning and NLP. I am also interested in Graphic Design and am proficient in Adobe Photoshop and Illustrator.

Work Experience

During my undergraduate life, I have also taken up various internships and the following is a brief summary of my previous internships:

• Dayta AI Ltd.

Machine Learning Engineer

• Used models such as BERT/ELMo to develop Chinese Paraphrasing tools
• Part of a 5 perso team responsible for developing a new state-of-the-art Chinese Tokenizer, which outperformed existing models.
• Worked on person recognition and headcount tracker projects, including doing transfer learning to fine-tune existing implementations
• Developed various in-house tools to create datasets to retrain computer vision models
• Completed many projects with a short turnaround time
• Assisted in building a news paraphrasing NLP system using state-of-the-art tokenization methods and neural networks
• First Code Academy

STEM Instructor

• Instructor for Intro to Python and MIT App Inventor courses
• Infotel Business Solutions Ltd.

Web Development Intern

• Received training in Full-Stack programming with emphasis on frontend development
• Assisted in debugging and quality testing of programs

Research and Publications

My research interests include NLP and Machine Learning! The following is a brief summary of paper(s) I've worked on:

• Learn Languages First and Then Convert: Towards Effective Simplified to Traditional Chinese Conversion

Pranav A, S.F. Hui, Ishaan Batra, I-Tsun Cheng and Chiu Yik Hei

Simplified Chinese to Traditional Chinese Character conversion is a common pre-processing step in Chinese NLP. Our proposed method achieved a near-perfect disambiguation accuracy of 99.98%, outperforming previous Chinese Character Conversion approaches by over 5 points in accuracy. Our paper was accepted at NAACL-SRW.

I am currently working on some more projects in this field at the moment so stay tuned for more!

Projects

I like to do fun stuff! Here are some projects I've worked on:

• COMP4461(Human-Computer Interaction) Project 1

Group Name: Basketcase

• Brainstorming Phase
Our group sat down and tried to list as many existing projects or topics that is related to sustainability projects. As a group that loves food, we decided to choose reducing foodwaste as our problem to tackle. Suprisingly, there were many projects which were school driven.

• Existing Projects
• In some Halls like TKO, there are food sharing cabinets where student could put preserved/canned intems and donate them to other people. This project however is limited to preserved goods, and doesn't show what items are currently on display.
• In LG7 and LG1 canteen, there are options to order with less rice with a 1 HKD discount. This however limits the option only to rice, and not on other items.
• HKUST also had an bring a lunchbox today campaign, but not everyone had the time or resource to bring a lunchbox to campus.
• Designs


After discussing about different ways to reduce food waste, we tried to focus on groups that cooked. As a student who regularly cooks, I always had the problem where it is very awkward to portion my groceries. We had several designs, and one that I worked on, was grosharies, an app that makes sharing groceries simple. Dohoon drew this storyboard, and we all went around to ask for opinions. The feedback had been overwhelmingly positive with only minor concerns.

• Grosharies

In the end, we chose the grosharies as our final idea as it had in our opinion was the most viable option, and had a far reaching potential meant for more than just UST students. We have agreed that the core functionality would be a intuitive interface to allow people with excess groceries to share, and buyers to locate and identify the goods quickly and easily. I implemented the app prototype in a week which great influenced the feedback and the other members thought it was brilliant!


• Reflection
During this course I learnt a lot about rapid prototyping. We had to come up with multiple designs really fast and weigh in quick on which design is the most viable option. All the ideas were equally enticing at first but by weighing them against each other we managed to choose what we thought was the best option. We also learnt that parallel prototyping works better in practice compared to serial protoyping in the course of our project. I also thought storyboarding was a fun process and definitely gave more insight into the design and cater better to needs. Due to the short timeline of the project it did feel a little overwhelming at times, however, I am really grateful to my teammates for making this a highly educational and fun project to do. It was certainly challenging coming up with a prototype in less than a week and we certainly did get our hands dirty doing that.
• COMP4461(Human-Computer Interaction) Project 2 : Augmented Reality

Group Name: Jockey Team

• Brainstorming Phase

Our first idea was adding AR figures or explanations to an educational book for studies but we thought we could find something more interesting. So we came up with AR business cards which can enhance creativity, the range of information, interactions, or help some people who actually use them for work to save time with shortcuts. It can actually make people use business cards for business and not stack them up without using them anymore. It would also create a way to make difference between candidates.

• Designs

Video Prototype
• Further Enhancements

We can add some options that you can find important, we want to create a lot of possibilites to have lots of different AR business cards. For example :

• Photos
• Shortcuts for messages; mails, phone numbers, linkedin pages, etc.
• Proves of work : websites, projects, etc.
• Games, designs, arts for each
• Dimensions Enhanced
• Range of Content Input : Adding information
• Information Exploration : Navigation through websites
• Information Findability : New shortcuts for phone numbers / messages
• Digitalization : Having more infos in less paper
• Analysis
• Our result is still a prototype but proves you can do everything with your own AR business card.
• “Wow” Factor When Looking for Jobs
• Can help people for business interactions BUT Accessibility has layers of complexion
• Ahead of its time
• Reflection

  Project 2 was definitely an interesting experience, since we got to experience an up and coming technology like AR. It was an enriching experience developing the prototype and making a proof-of-concept with my teammates in a limited time and it really opened up our eyes to the immense potential of this technology. I would definitely love to undertake more projects in this field in the future. I am grateful for my team for putting in effort despite our other project commitments.

• COMP4461(Human-Computer Interaction) Project 3 : Human-Robot Interaction

Group Name: N/A

• Introduction

The purpose of this project was to identify one scenario where a physical robot could be of service in education, as a tutor, companion, assistant, peer, or student. Our group sought to understand the need to use a robot interface, and where a robot would be more useful than other solutions.

• Brainstorming
Our group created a mind map based on areas in life that often need a teacher. We thought of various times when we (or other people we have observed) would have liked to have a robot to help us increase knowledge and skills, such as for language or vocabulary learning or music learning.


Design

Here is the video of some members of our group acting out our robot design.

Storyboarding

We created four storyboards to represent our top solutions. We used the storyboards to ask users which one they liked the most and why. We used the feedback to help us choose our final design. Language Translator and Teacher Wikipedia / Fun Fact Bot Music Instructor Robot Vocabulary Teacher Bot

Speed Dating

Some feedback we got about our storyboards was that the language, Wikipedia, and vocabulary bots were too simplistic of tasks that did not really require a robot. The music instructor bot seemed to be the most creative and unique option for interviewees. It had the largest array of capabilities that could aid in music learning. Suggestions from interviewees included having the robot detect improper rhythm and tempo, false notes, or if the musician was out of tune; suggest music to play based on the user's skill level; or even play a metronome.

Constructing POV

Based on user feedback, we chose the music robot idea, which we called Beat Bot, and constructed the following POV.

User:
• People who play musical instruments or sing
• People who don’t have a real music teacher to get feedback from
• People who prefer to practice alone
Needs:
• To practice musical skills in their own time
• To have their music errors be detected in real time and get feedback
• To have a fun, interactive musical buddy
Insights:
• Allows people to take authority in their own musical learning
• Helps beginners who have anxiety with learning new things or practicing in front of other people

Choreographe Code

The following is the Choreographe code we created. Our team was unable to test the robot in the lab, but we created a video of us acting out how the robot would work.

Personal Contribution

During the brainstorming phase I pitched some ideas and eventually supported the our final idea based on my previous experince as someone who once tried to learn the piano. Furthermore, I worked with Megan and we conducted the speed dating interviews. I assisted in adding choreographe blocks and also helped construct the POV. I was also responsible for the creative direction of the video.

Reflection

Getting hands on experience with the choreograph software was definitely one of my favourite parts about this course. I believe in the future use of robots in the musical field will be widespread, however when evaluating subjectively I believe humans still hold the edge due to our ability to understand and enjoy creative expression. Furthermore, robots can help master basics in the absence of a human teacher since they can accurately pinpoint mistakes and help people hold accountability and prevent errors from being embedded. While I feel, that there may be an initial adjustement period since the entire experience might be a bit awkward at first, but getting over this initial period will eventually yield better musicians in the future.

My email address is ibatra@ust.hk.