Projects and Research
I have worked on numerous projects at work, for research, in classes and during hackathons, and have always tried to solve real world problems to impact the lives of everyday people through purpose driven product design. I thrive in work environments that require lots of teamwork and collaboration, and strive to set high standards through the quality of my work.
Warehouse Inventory Traceability Robot
Vimaan Robotics
Feb 2022 – May 2022
Headed the development of a new forklift-based AI system (PickTrack) to improve traceability of warehouse inventory.
Learn More
Smart Mailbox
Columbia University
Oct 2021 – Dec 2021
Misdirected mail is a common issue, often requiring homeowners to manually handle or forward items meant for others. The Smart Mailbox solves this by automatically scanning envelope text to determine whether mail is addressed correctly. If valid, it accepts the mail, notifies the user via text, categorizes it (e.g., spam or important), and tracks trends. If not, it rejects the mail and alerts the deliverer to return it to the sender.
The ICM is a consumer-facing device that not only improves efficiency, but also helps secure your mail by providing real-time notifications when new mail arrives. Designed for individual households, especially in suburban or rural areas, it also supports users with disabilities by simplifying mail handling and reducing manual effort.
View our project on GitHub
| Feature | Description |
|---|---|
| 1) Mail Detection | Detect that a new mail was inserted into the mailbox and click a picture |
| 2) Optical Character Recognition (OCR) | Process the image using openCV and read the text on the envelope |
| 3) Correct Delivery Checker | Determine if receiver name on the envelope matches the name of the mailbox owner, and returns envelope to mailman if not |
| 4) Mail Categorizer | If mail was accepted, use ML model to categorize and label mail as Spam / Not Spam |
| 5) Smartphone Notification | If mail is not spam, then send an SMS to user's smartphone with sender's name |
| 6) Data Visualizer | Show data trends over time such as percent of spam mail, time of day mail is delivered, and number of mails currently in mailbox |
Components
Camera Web Server
OCR Locations
Number of mails per day
Robot Metabolism Project
Columbia University - Creative Machines Lab
Feb 2021 – May 2021
Research Assistant with Dr. Hod Lipson.
The Robot Metabolism Project is a modular robotics platform that can self-assemble, self-replicate, self-repair and self-improve. We developed a platform that can identify, harvest and integrate robot components to better itself. This work will push the boundaries of artificial life and modular robotics research. Our link has two actuators with an embedded optical sensor to track its absolute position in space. We built an interaction system using a server infrastructure and ran experiments to characterize the performance of our links.
• Manufactured and assembled hardware (servos and microcontrollers) of modular robot links (building blocks).
• Identified design improvements on Fusion360 and implemented control algorithms on Photon using C++ to ensure links moved precisely.
• Co-authored research paper that introduces framework for self-sustaining robotic ecosystems.
Published in Science Advances | Read Research Paper Here
MATE ROV
Columbia Robotics Club
Sep 2020 – Apr 2021
• Created a virtual model (digital twin) of the rover and optimized design using SolidWorks.
• Simulated robot motion and control in an underwater environment on ROS gazebo using lift-drag plugin.
Next steps:
• Integrate computer simulation to a physical underwater robot.
• Write scripts in Python to implement publisher/listener methods that print the location, orientation and status of a single node.
Automated Antenna Tester
Persistent Systems
Jun 2021 – Sep 2021
• Conceptualized, designed and built 3 test fixtures to automate the testing process of an antenna power supply.
• Applied technical knowledge of Hardware and software to construct a product that autonomously evaluates the functionality and characterizes the performance of the device-under-test.
• Wrote Python scripts on a Raspberry Pi to control power relays in an embedded system and perform data throughput tests.
• Developed Bash-shell scripts to automate the configuration of network settings and setup of Windows and Linux machines.
Produced and Donated Personal Protective Equipment
In early April 2020, as the coronavirus outbreak in New York City hurtled toward full-blown crisis, I volunteered my skills in engineering design and 3D printing to support the manufacturing of personal protective equipment (PPE) which were in short supply. Like many members of the 3D printing circle keen to assist in slowing the spread of the COVID-19 pandemic, I got to work in my apartment producing prototype face masks and face shields by modifying open-source designs. What started as a hobby, quickly turned to a duty to help frontline workers and others leading the response effort in whatever small way I could.
If you know someone with access to a 3D printer, please join the movement! Use the contact form on the main page to request the 3D files.
Order Form
Dentist Face Shields
Personal Project
May 2020
In the peak of COVID, a local dentist reached out to me saying that her clinic was in dire need of face shields, as the ones they ordered online were delayed and new orders would take too long arrive.
Problem Statement:
• Design a face shield to be worn over the face of dental professionals
• Reduce the risk of contact with infectious splash, splatter and spray to protect face better than glasses and masks alone
Constraints:
• Distance between face and the shield must be 2.5" at minimum, because magnifying loupes and lights have to fit
• Shield must extend as far back as possible from the sides, preferably up to ears
• Comfortable to be worn for a long period of time
Results:
• Plastic covers the full face, up to the ears, and fits over glasses, magnifying loupes, lights and masks
• Cushion lining between forehead and mask allowed for a comfortable, relaxed fit
• Ability to adjust length of elastic strap by untying the knot
• Shields can be disinfected, washed and reused, thus reducing waste
• All parts of the face shield can be individually replaced
Front View
Back View
Dentist 1
Dentist 2
COVID Data Visualizer
Personal Project
Mar 2020 - May 2020
In 2020, we were flooded with all sorts of COVID-19 related data. While it was important to stay up to date with the latest information in the news, it was also important to understand the data we were looking at to make better, more informed decisions. As states started to reopen in late 2020, the perception and understanding of the lawmakers and general public had to align with the facts.
Click the visualizations below to interact with the full dashboard. Get the data: OWID
Python Visualizer | Tableau Visualizer
Machine Learning: Gesture Recognition
Columbia University
Oct 2021 - Nov 2021
• Cloud Server: Set up a cloud server on Amazon Web Services (AWS) and send continuous accelerometer data of at least 10Hz. Display the data on the cloud through the terminal
• Databases: Created a MongoDB database (with key-value pairs) to be hosted on the AWS server and we will access it using HTTP commands. This database will be used to store accelerometer sensor readings from our smartwatch and was used as our labeled training data
• Supervised Learning: Trained a Random Forest Classifier model (using Scikit-learn) which was capable of recognizing gestures based on accelerometer data from the smartwatch and classifying the letters “C”, “O”, “L”, “U”, “M”, “B”, “I”, “A”
• Visualization: Deployed the model onto the AWS server where it enabled real-time classiï¬cation of the letters. Display the word “COLUMBIA” one letter at a time
Smart Bin Remover
Purdue University
Jan 2016 – May 2016
For our Senior Design Capstone Project, my team and I designed and built an innovative, fully autonomous delivery robot that can attach to the three most common types of waste bins and deliver the attached bin to a desired pick up location on trash day. I headed the programming of an Arduino microcontroller in C and integrated various sensors and “smart” technologies to remove all user interaction from the feedback control loop.
First, we defined the problem we were trying to solve, then we built a prototype mechanical system, tested the software and controls, and evaluated the results. We repeated these steps numerous times and, several iterations later, ended with a product that was completely different from what we had imagined. We looked at various aspects of the robot, such as ease-of-use for the end user, durability of our product, ability to integrate with the existing systems and ability to meet the design parameters set by our end user (the local Waste Management company). We also designed the charging pod and programmed the robot to find its way back to the base, where it would plug in and charge itself when not in use.
The robot is an artificially intelligent rational agent designed to safely drive waste to collection point and back to home base.
• Performance Measure: safe, fast, avoid bumps so waste doesn't spill
• Environment: roads, driveway, pedestrians, weather
• Actuators: DC motors (future: steering, brakes, turn signals, horn, etc)
• Sensors: date and time, ultrasonic and infrared proximity sensors, line following reflectance sensor (future: load cell to measure weight and tell if waste has been collected, cameras to avoid obstacles, GPS , speedometer, accelerometer, odometer, etc)
The Smart Bin Remover project placed 7th place (out of 49 total submissions) at the Mechanical Engineering Innovation Competition at Purdue University.
Watch Video | Pitch Deck | Photo Gallery
Project LiT
University of Illinois Urbana-Champaign - HackIllinois Hackathon
Feb 2016
Received recognition for creating one of the best hardware hacks at Hack Illinois, one of the largest University Hackathons in the country with over 1000 participants. We created an internet connected wearable breathalyzer to warn people when they are over the alcohol permissible limit for driving. It uses an OLED panel that notifies you of your blood alcohol content and an RGB LED strip that alerts everyone around you if have crossed the legal limit, so your friends can stop you from driving or stop themselves from getting in the car with you. My contributions included implementing the engineering design process and writing code to integrate all the electrical components using an Arduino microcontroller. Project LiT won top 5 for best hardware hack out of 216 total submissions at the HackIllinois Hackathon at University of Illinois at Urbana-Champaign.
Project LiT | Pitch Deck | Photo Gallery
Project Green Light
Purdue University - BoilerMake Hackathon
Oct 2015
Project Green Light is a social initiative which helps people connect more easily. We struggle and often find it awkward to approach and meet new people in the real world. Over the weekend-long hackathon, my team and I wanted to design something new and bring it to life, and we believed that our product would revolutionize social interactions. We wanted to bring the comfort of interacting with people behind a screen to the real world, using a simple and unobtrusive system. We tried to solve the social engineering problem of finding seats in crowded locations with tables, by building an LED box with switches for occupied, silence or mingle statuses for each table. The beacon provides users with information such as table availability and status through data collected from the Arduino devices. In addition, we want to create a tool which can measure the rate of change in the willingness for people to socially interact with one another in any given environment. My contributions included collaborating on the design and build of the product, developing the product all the way from ideation to completion and working on the presentation and documentation for the project. Project Green Light won top 5 for best hack out of 84 total submissions at the BoilerMake Hackathon at Purdue University.
Project Green Light | Pitch Deck | Photo Gallery
T2P2D - Toys to Physics to Design
Purdue University - Convergence Design Lab
Aug 2014 – Dec 2014
Research Assistant with Dr. Karthik Ramani.
The purpose of this project was to develop a framework to encourage a maker-based design mindset embedded in physics for K-12 students.
• Developed STEM-focused teaching "Engineering by Design" workshops for K-12 students to design simple toys and 3D print them.
• Benchmarked different open source CAD software and categorized them by ease-of-use.
• Manufactured prototypes using 3D printing technologies (Ultimaker 2, Makerbot).
HandiMate
Purdue University - Convergence Design Lab
Aug 2014 – Dec 2014
Research Assistant with Dr. Karthik Ramani.
Robotics serves education in many ways. The process of developing robotic solutions provides a rich and meaningful context for engaging students. In this project, we developed a modular robotics kit to animate crafted toys.
• Created a robotics kit which enables users to construct and animate their toys using everyday craft materials.
• Proposed new and innovative uses for the product using everyday objects such as cardboard, construction paper and spoons.
• Tested and gave feedback on various constructions and promoted brainstorming among users.
