CSE 2024 Spring Departmental Demo Day

picture of demo day participants

We're thrilled

to invite you to the thirteenth bi-annual Comp. Sci. & Eng. Fall Demo Day. Student groups from several CSE capstone classes will be presenting the culmination of 3-months of effort, hard work, (metaphorical) blood, sweat (well... caffeine really), and tears (see above).

  Where:

Davis Hall; 1st and 2nd Floor Atrium

  When:

Tuesday, May 7th, 2024

Schedule

2:00 PM - Event opens
2:05 PM - Opening remarks
2:10 PM - Judging starts
4:00 PM - Judges will get together to review results and pick winners
4:30 PM - Winners announced and videos played for the winning teams
5:00 PM - Event closes

Setup

Setup will start at Noon. Tables will be available for both sponsors (each sponsor will get a table) and demo participants (2 to a table). Easels will be available for participants. If you need power, please let us know! If you have any other special requests, please contact ahunt@buffalo.edu to let me know, and we will do our best to accomodate you. There are two hours reserved for setup - you can come at any time during that period to get organized, but please make sure you leave yourself enough time to be ready to go by 2PM, to give you the chance to network.

Networking

Before we open the atrium to students and the public, we’ll have some time reserved for the participants to come and chat with the sponsors and the judges. Pizza will be there as well (A big thank you to our sponsors!), so that the participants and sponsors can have a chance to eat before demos begin!

Judging

During the demo, judges will circulate to the participants demo stations, and they will be rating each project on a specific set of criteria. Judges, expect to spend approximately five minutes with each team, in order to give you time to see them all. You will be assigned a set of projects to view specifically, but you can feel free to talk to more teams as time permits! Teams, keep this in mind and keep your presentations crisp and to the point!

Prizes

There will be prizes for the top teams selected by the judges. They will be announced in the atrium, and there will be a quick photo op for each winner. Good luck to everyone, and I can’t wait to see you all there!

Winners

1st Place - SAARTHI

2nd Place - Signify

3rd Place - Youro

ACV Auctions Innovation Award - SAARTHI

Acknowledgements

Sponsors

Judges

  • Chris Miller
  • Anarghya Das
  • Amir Nassereldine
  • Ethan Blanton
  • Jim Brandt
  • Haonan Lu
  • Kyle Daving
  • Karlene Kardysauskas
  • John Tantillo
  • Sonny Sonnenstein
  • Neel Tripathi
  • Hannah Wilcox
  • Joseph Forsyth
  • Will Giegerich
  • Kyle Shuttleworth
  • Nicholas Myers
  • Alina Vereshchaka

Presented Projects

(CSE 700)

"Code-Free Data Science Pipeline: Tailored Machine Learning Models with Automated Data Preprocessing Platform"
  • Chennuri Ullas Reddy
  • Boddu Srivatsa
"Experience the 'Code-Free Data Science Pipeline,' a user-friendly platform tailored for beginners and seasoned professionals alike. Simplifying machine learning model creation and data preprocessing, our intuitive interface guides users through every step. With automated data cleansing and transformation, users can prepare datasets effortlessly. The platform's standout feature is its ability to generate tailored models based on user preferences and dataset characteristics. Say goodbye to manual coding and debugging—unlock valuable insights with ease. Whether you're diving into data science or streamlining your workflow, our platform empowers you to harness the power of machine learning effortlessly."
Crime Dynamics Mapping: Textual, Spatial and Temporal Perspectives
  • Tarun Reddi
  • Charvi Kusuma
This research employs advanced machine learning and neural network models to analyze crime types, predict hotspots, and uncover latent topics within Los Angeles crime data. By integrating spatial, temporal, and textual analysis, it offers valuable insights for law enforcement on crime patterns and prevention strategies.

(CSE 370: HCI)

200 OK
  • Adavita Jain
  • Kevin Zhong
  • Michelle Ye
  • Neem Zaman
  • Jason Zheng
A social media platform for gamers all alike to post, rank, and find each other based on game that they play.
UI maxxing
  • Andrew O'Lansen
  • Tiffany Cai
  • Jessica Mui
  • Journey Spratt
  • Sheba DeLozier
Recipe Pantry is a social media application designed to assist those looking to share, receive, and search for recipes. Our app makes usability the utmost priority for the end of streamlining users towards discovering the recipes they're going to be most interested in. The flames fueling the fire behind Recipe Pantry cook in Recipe Creation and Cookbook Curation. The former allows one to title a recipe, describe the ingredients and preparation steps therein, and to accompany it with an appetizing image. Cookbooks allow for the collation of recipes into a cohesive whole - say a collection of Latin cuisine, or perhaps Asian Fusion? Recipe Pantry grants the whole world of cuisine at an instant!
Won One
  • Devon Morlok
  • Ibrahim Allahbukhsh
  • John Abramo
  • Robert Reyes
  • Smaran Vedantam
  • Owen Jin
Trek is a place where nature lovers will be able to gather to plan and share hiking trips. Users can engage with two main sections: the gallery page and the hikes page. The gallery page is a feed that allows users to post pictures/descriptions relating to hikes they are on/have been on, and other users can interact with these posts (like/comment). The hikes page is a feed dedicated to posting/planning hikes. On this page, posts would include the location where the user is planning to take the hike, a description, tags that specify which type of hike is it, and the level of difficulty. Other users can like, and comment on these posts as well as join the hike and will be put into a group chat to communicate with all hikers part of the hike. Users have the choice to restrict the visibility of their hike and gallery posts to friends to protect their privacy and limit people who can join their hike to friends. Once the hike is over the post owner can mark it as completed and the post will no longer show up on the hikes page. Users can see all of their completed and upcoming hikes on their profile page alongside stats like total miles hiked, and total hikes completed.

(CSE 442: Software Engineering)

Backend Boys
  • Alex Tisdale
  • Alp Buz
  • Brendan O'Connell
  • Matt Collopy
  • Marcus Hartman
Transforming Retail with Smart Self-Checkout: Revolutionizing shopping experiences through innovative, user-friendly kiosks for seamless transactions and elevated efficiency.
Diet Coke
  • Eduardo Turcios
  • Chris Dearing
  • Alexander Gherardi
  • James Lee
  • Jontan Khuu
An RPG based around the day-to-day experiences of a computer science major with fantastical elements intermixed.
InSight
  • Jiewenhu
  • swu65
  • Yzhu53
  • iallahbu
UB professor ranking and reviewing web app
Power Chess
  • member
  • member
  • member
  • member
  • member
Multi and single player Chinese chess web app
SERVS
  • Shreyas Narayanan Sridhar
  • Vrushaali Nagaraj
  • Eleanora Undrus
  • Sebastien Bowen
  • Ria Gupta
A website that blends the repository sharing capabilities of GitHub with the interconnectedness of LinkedIn.
The Climb
  • Donovan Blount
  • David Reilly
  • Aidan O'Donnell
  • Nikkos Rose
  • Annusha Pervez
A centralized, accessible site for users to log their exercise and nutrition habits all in one place.
The Fortniters
  • Braeden Cairns
  • Edwin Irizarry
  • Navaneeth Nakka
  • Jason Nguyen
  • Justin Podbielak
The project is a fun and interactive 2D Dungeon-Style single-player game that allows the user to add computer enemies to their party to play alongside them.
The Journalists
  • Rudraaksh Karthick Bhuvaneswari
  • Beibit Zhylkaidar
  • Karthik Khatri
  • Nick Sukhdeo
Journalist is a dynamic online journaling platform tailored for University of Buffalo students.

(312)

BlueBrilliantv2
  • dylanzin
  • lgrichar
  • masonpug
A full stack single page webapp with full multi-player, player vs engine, and a forum for updates on everything chess related!

(CSE 302/303)

Choreographic Programming
  • Alex Kim
  • Tiffany Cai
  • Vincent Chan
Choreographic programming aims to simplify distributed systems / concurrent systems programming by allowing a programmer to specify the system in its entirety
EL/R - Indoor Wi-Fi based Localization and Navigation App
  • Anthony Szykowny
  • Christopher Varghese
  • Matthew Cheung
  • Dora Lei
  • Sydney Chan
  • Ashton Hannon
Indoor Wireless Localization (specifically Wi-Fi) has become quite widespread, there is an active Task Group 802.11bf as well to bring this into standards and there are many learning based solutions to bring localization accuracy to sub-meter level. While, this is great news for indoor automation, there are deployability concerns in terms of generalizability and privacy concerns surrounding it. While, the most popular approach is to use data based ML-models to provide sub-meter accurate localization algorithms, there are still requirements for generalization to different environments and better accuracy. We developed a web-app to get data from Wi-Fi routers for a given smartphone. The data is then to be processed online to pass through an ML-model. We also developed an android app that can allow us to navigate within the first and third floors of the Davis Hall.
pgAQP Plugin
  • Winston Peng
  • Yudian Ke
  • Marcus Guild
Porting a B-Tree database structure from being a postgres plugin, to being a standalone C library that can be plugged-in to any relational databse.
Teamwork Tool
  • Ardian
  • Marian
  • Tariq
  • Matt
The Teamwork Tool project exists for UB faculty to distribute peer evaluations among their students. Many UB courses require group work and it is important for these classes to have a peer evaluation survey to weigh individual grades properly. Many current online form platforms, like Google Forms, lack services that allow grades to be adjusted according to survey responses. Our tool will have the functionality of grade weight. This tool is also tailored to UB and will automatically collect student information, like name, UBIT name, email, which is something current online platforms lack. Our tool will also allow rubrics to be shared among other professors to reduce unnecessary form duplications. To conclude, this tool aids in the process of creating, distributing, and collecting the surveys for professors.
Virtual Ampitheater
  • Brian Ren
  • Christos Gogos
  • Phil Giambrone
  • Vincent Lin
To design and develop a VR visualization of a newly discovered amphitheater by Professor Sandro for an introductory class for archeology. The program should be historically accurate as to allow students to gain a better understanding of the site by interacting and analyzing it.
Vizier: Data Visualization Notebook
  • Cameron Toy
  • Lawrence Yao
The Vizier team is enhancing computational notebooks with a user-friendly data visualization component that simplifies creating sophisticated charts from datasets. This initiative, distinct for its ease of use and seamless integration with Vizier’s existing features, allows for direct manual adjustments via Vega, significantly improving flexibility and usability. Vizier wants to take this opportunity to develop a simpler, but more powerful data visualization component.
Java and Android Malware Detection Using SootUp
  • Ria Gupta
  • Vrushaali Nagaraj
  • Shreyas Sridhar
  • Jackey Wang
  • Nawar Khouri
Our project aims to design an interface that takes in Android or Java source code and outputs any detected leaks of sensitive information. We use visual debugging to represent the worklist algorithm that traverses a piece of code and detects if it is malware or not.
DevU
  • Keifer Myers
  • Saiwang Xiang
  • Kevin Zhong
  • Rishi Sahu
An extensible code autograding web service built as an alternative to Autolab. Built for students and instructors alike, DevU offers a simple yet powerful way to manage and automatically grade programming assignments.

(CSE 611: Master's Capstone Project)

CSE611 Kaleida Health
  • Mokshita Gupta
  • Harrison Hutton
  • Christopher Dushkoff
  • Jennifer Tsang
  • Gursimrat Tiwana
This project presents the redone interfaces with augmented backend support. In the new website, patients can share their medical journeys, see diagnoses and info assigned to them by doctors, etc. Doctors and admins can use their respective views to edit, assign, delete, and more on patients, specialities, medical centres.
Find a Mechanic
  • Tyler Landivar
  • Sonam Barnala
  • Sai Ranganadha Nisanth Chilakamarri
  • Rohith Merugu
  • Saisree Peddireddy
Find a Mechanic is an application that allows users to book appointments with local automotive service centers and lets them exchange messages with these garages to ensure complete transparency regarding the status of their appointments. Service Centers may sign up for a different type of account that will make them discoverable by users, and allows them to manage employees, garage information, and appointments within the app.
Look The Part
  • Narsimha Sashank Malayanur
  • Lakku Prathap Reddy
  • Vamsi Kumar Naidu Pallapothula
  • Hari Preetham Reddy Nandyala
  • Sai Venkata Swetha Gadde
This fashion app allows enthusiasts to purchase apparel and accessories inspired by TV shows and celebrities, offering an interactive platform to buy outfits that match those seen on their favorite stars. Built with React Native, it supports both Android and iOS devices, providing a seamless and intuitive user experience. The highly responsive design enhances visual appeal and user engagement. The app features optimized state management using Redux, reducing redundant data fetching and ensuring efficient performance during complex interactions within the dynamic fashion marketplace. Combining cutting-edge technology with user-centric design, this app merges fashion with visual media for a streamlined and engaging shopping experience.
RedPrint - Fitness
  • Chaitanya
  • Vijay
  • Surya
  • Priyanka
  • Poojitha
Redprint is a robust and user-friendly platform designed to streamline gym management, enhance user relations, and facilitate organizational management. It serves as a one-stop solution for gym owners, staff, trainers, and administrators, offering a range of features tailored to each user’s role and access level. From a business perspective, Redprint serves as a valuable tool for managing gym operations, maintaining user relations, and ensuring smooth organizational management. It allows for efficient delegation of responsibilities based on user roles and enhances the overall productivity of the gym management process. From a user perspective, Redprint offers a seamless and intuitive user experience. Each user role has access to features that are relevant and necessary for their responsibilities, ensuring a streamlined and efficient workflow.
Youro
  • Sailesh Reddy Sirigireddy
  • Mounikananda Reddy Kanuparthi
  • Abhishek Katta
  • Pranay Bunari
  • Santhosh Pothuganti
Youro is a healthcare web application that enables patients to consult qualified doctors for a wide range of urological issues and conditions virtually. For doctors, this application provides a comprehensive dashboard with patient information, upcoming appointments, and more. It offers patients a robust user experience, allowing them to easily book appointments, access their prescribed care plans, obtain detailed information on each prescription, and engage in quick chats with doctors. Additionally, this application empowers patients with additional tools for self-assessment and understanding of their urological health. Users can take a tailored questionnaire designed to gather relevant information about their symptoms, medical history, and lifestyle factors. This questionnaire aids in providing a quick initial diagnosis. This application also helps doctors better understand the patient's condition before the consultation by summarising the patient responses to these questionnaire. Innovatively engineered with Spring Boot and ReactJS technologies, this application seamlessly integrates state-of-the-art solutions with user-friendly design principles. Additionally, the chat functionality, built with WebSockets, ensures real-time communication, enhancing the immediacy and effectiveness of patient-doctor interactions. By harmonizing healthcare with convenience, this application delivers a fluid and engaging experience, fundamentally transforming the patient-doctor interaction landscape in urological care.

(CSE 498)

Efficient Signature-Based Intrusion Detection Systems
  • Matt Kreuzer
An exploration of Intrusion Detection Systems (IDS) and how they can efficiently perform their tasks to detect malicious packets. In particular, signature-based detection is implemented in Go by experimenting with a variety of string-matching algorithms for performance (Boyer-Moore, Rabin-Karp, Knuth-Morris-Pratt, and Aho-Corasick). Additionally, this project explores how an IDS would communicate to a Security Information and Event Management (SIEM) system once a match was found. A simplistic SIEM web interface with a single IDS sensor (attached to a vulnerable Apache machine) is implemented to perform this task, using syslog-ng for communication.

(CSE 635: Natural Language Processing )

Elective Genie
  • Lalla
This project developed a Retrieval-Augmented Generative AI to assist University at Buffalo students with academic planning. It leveraged LangChain, Pinecone, and a sentence transformer model to encode course data into semantic vectors, enabling efficient semantic search. The top retrieved results were fed into Hugging Face's Llama-2-7b LLM to generate comprehensive natural language responses.

(CSE 468/568 Robotics Algorithms )

F1tenth Racing Simulation
  • Pranay Meshram
  • Christo Aluckal
  • Yashom Dighe
  • Yash Turkar
Join us for a live demonstration featuring the top performers from our final class assignment, where students have developed an end-to-end racing pipeline using the ROS Algorithm within our custom simulator. This highlights the advanced techniques and considerable effort invested by the students to create a comprehensive autonomous racing system. Witness firsthand the innovative solutions that propel these autonomous vehicles across the digital racetrack.

(CSE 700 Independent Study)

It is Not True that Transformers are Inductive Learners
  • Michael Sullivan
I show experimentally that: (i) transformer NLI models treat external negation as a distractor, and effectively ignore its presence in hypothesis sentences; (ii) several near-SoTA models fail to inductively learn the law of the excluded middle (LEM) for a single external negation prefix with respect to NLI tasks, despite extensive fine-tuning; and (iii) those models which do learn LEM for a single prefix cannot generalize this pattern to similar prefixes. However, I then prove theoretically that there does exist a transformer architecture capable of modeling LEM, suggesting that current training procedures and/or the structure (i.e. lack thereof) of purely textual data may be insufficient for transformers to inductively learn LEM.

(CSE 700 )

WealthWise
  • Lakshmi Priyanka
Financial Management and Recommendation platform.

(CSE 676: Deep Learning)

HoloZipper: A Self Supervised Learning based RGB-D Image Compression Algorithm
  • Debosmit Neogi
  • Shivan Mathur
Compression of RGB-D data is a challenging and important task because uncompressed RGB-D data can be extremely slow to transmit over the transmission channel. AR-VR applications rely on fast and optimized streaming of large amounts of RGB-D data. Hence for optimized bandwidth utilization, compression is required. The challenge here is to gain a high compression ratio as well as have a high quality reconstruction of the compressed images at the users’ end. Presenting HoloZipper, an end-to-end trainable self-supervised learning model architecture that can achieve a compression ratio of 64 on RGB-D data. HoloZipper consists of two separate encoder-decoder architectures that take as input RGB and Depth data respectively. The encoder returns a compressed representation of the data and feeds it into the decoder architecture. The decoder then reconstructs the compressed data into the original data with an average PSNR of 28. Unlike typical autoencoder architectures, HoloZipper uses discrete encoder-decoder models to better serve real world applications.
SAARTHI
  • Chirayu Sanghvi
SAARTHI (Smart Auto Assessment & Roadside Technical Help Interface) is a cutting-edge web application designed to revolutionize how car owners manage vehicle damage. This innovative platform utilizes advance AI technologies, including object detection, instance segmentation, and salient object detection. It also swiftly and accurately assess damage from user-uploaded image. Upon assessment, SAARTHI classifies the damages into major or minor categories and provides an immediate repair cost estimate. This seamless integration of AI not only allows users to receive real-time damage evaluations but also enables them to interact directly with our chatbot for various roadside assistance services, such as calling for instant help, on-the-spot repairs, or towing. Each user creates a personal account on our portal, ensuring a personalized and secure experience. Once damage is reported and assessed, users can engage with the chatbot to generate detailed assessment reports and initiate contact with nearby affiliated repair shops for prompt service, all through the SAARTHI application. Beyond its technical capabilities, SAARTHI offers significant benefits by enhancing transparency and efficiency in the automotive repair and insurance sectors. Users are protected from inflated charges with a trustworthy, independent assessment, aiding them in navigating the complexities of car repair and insurance claims. Additionally, the system streamlines operations for insurance companies by reducing reliance on manual inspections, which can be time-consuming and occasionally inaccurate. SAARTHI also promotes social well-being by ensuring fair transactions and fostering trust in automotive services, ultimately contributing positively to societal equity. This project demonstrates its potential to transform the automotive service industry through technology-driven solutions and a customer-centric design.
Eyes on Eats: From Image to Formula
  • Tarun Reddi
  • Charvi Kusuma
Eyes on Eats is a project that leverages deep learning for object detection and text generation to analyze images of ingredients and create personalized recipes, reducing food and financial waste, saving time on meal planning, and making cooking accessible to everyone.

(CSE 676 : Deep Learning)

ADAS (Advanced Driver Assistance System)
  • Sai Deep Muvva
  • Sachin kumar Koppula
  • Kodakandla Srija
ADAS is a deep learning model for recognizing traffic sign board images. The problem revolves around the need for efficient and accurate recognition of traffic signs in real-time scenarios. With the increasing prevalence of autonomous vehicles, robust traffic sign recognition is crucial for enhancing road safety and improving navigation systems. Also, equipping non-autonomous vehicles with this system to assist drivers to improve the quality of driving and ensuring safety. By leveraging deep learning techniques, this project seeks to address the challenges associated with varied lighting conditions, occlusions, and diverse sign designs, making it an interesting endeavor in the realm of deep learning and transportation technology.

(CSE 546: Reinforcement Learning)

Double DHADEL - MARL for Cricket
  • Anirudh Jayanthi Rama Sesharka
  • Karthik Pavan Chowdary Bobba
  • Garlapati Sushma Reddy
This project aims to train cricket batting and bowling agents to operate within an adversarial Multi-Agent Reinforcement Learning (MARL) environment, where they compete to gain rewards against each other. The batting agent focuses on scoring runs efficiently, while the bowling agent strives to take wickets and restrict runs. By simulating competitive scenarios, the agents learn strategic decision-making, ultimately enhancing their performance in real-game situations. We developed the innovative Decomposed Heuristic Advantage-Driven Efficient Learning(DHADEL) technique by decomposing complex tasks into manageable actions and rewarding each phase based on advantages gained. This approach enables agents to effectively learn intricate skills in cricket batting and bowling, leading to improved performance through strategic decision-making.

(CSE 676 B: Deep Learning)

Signify - Web Application for ASL detection using deep learning model
  • Manasi Jadhav
  • Sneha Singh
  • Sneha Yadav
Signify is an interactive web application designed to bridge the communication gap between ASL users and non-users by providing real-time detection and translation of American Sign Language (ASL) using deep learning technology. This application aims to bridge the communication gap between ASL users and those who do not understand ASL, facilitating smoother and more effective interactions.

(Deep learning)

Deep fake image classification
  • Mahitha balumuri
s

(Would like to showcase what i have been working on)

Kai
  • Raj Khatik
  • Nalini Ratha
The project demonstrates implementation of multi modal LLM on humanoid robot pepper to understand human speech and answer question based on what it sees through computer vision and vision models. Prime goal of this project is to showcase AI deployed on robots in real world scenario

(CSE 676)

Enhancing Educational Accessibility
  • Jayakrishna Gandhamalla
  • Anip Kumar Paul
  • Ashith Nithyananda
We built an automated system that transcribes live lectures using Speech to Text API. It also generates concise summaries using our trained summarization model, for every specified minutes. Both live transcription and generated summary is displayed to the listeners in the real time. We have also provided the interface to join and create the meeting.

(CSE799)

Human robot interaction with Nao, Pepper and Unitree Go2
  • Yadvender Singh
  • Sunil Rufus Ramneedee Pushparaj
Pepper- A robot assistant who can answer all university related queries of students and general queries using Language models. Nao- A small friendly Humanoid robot who can imitate movements of people infront of it and alert it detects someone falling. Unitree Go2- A quadruped that can listen to voice commands, follow people around and detect unidentified objects.

Past Demo Days

Fall 2016 Fall 2017 Fall 2018 Fall 2019 Spring 2019 Fall 2020 Fall 2021 Spring 2021 Fall 2022 Spring 2022 Fall 2023 Spring 2023 Spring 2024