Cyclone
Eye of the Cyclone | Analysis of Tropical Cyclone Genesis
This project studies how well the AI weather model Pangu-Weather predicts tropical cyclone genesis compared with ERA5, a physics-based atmospheric reanalysis used as the reference baseline. It focuses on identifying where the AI model systematically differs from ERA5 at the moment storms begin to form, especially in key storm characteristics such as pressure, wind, asymmetry, and warm-core structure. The team analyzes storms from 2018 to 2024 using feature extraction, principal component analysis, clustering, and outlier detection to group forecast errors into interpretable failure modes. The overall goal is to understand when the AI model performs well, where it struggles, and what improvements would make its tropical cyclone forecasts more reliable.
Student Team Members
- Aaron Wu
- Tri Hoang
- Youssef Gehad
- Dori Olson
- Mason Weiss
Sponsor and Mentors
- Sponsor Mentor: Prof. Avantika Gori
- D2K Fellow, PhD Mentor: Kooshan Amini
- Faculty Mentor: Prof. Xinjie Lan
Double Bass
The Bass Analysts | Double Bass Analysis
This project analyzes antique double basses from the 17th century. The double bass is significantly understudied compared to other string instruments like the violin. This is in large part due to the large size of the instrument. We analyzed three dimensions of the instrument: air volume, f-holes, and wood thickness. Studying these components can help us better understand how antique basses were constructed and inform the construction of future basses.
Student Team Members
- Eric Ching
- Justin George
- Isabel Wang
- Camille Wong
- Haotian Zheng
Sponsor and Mentors
- Sponsor Mentor: Timothy Pitts, Mark Kindig, Duane Rosengard
- D2K Fellow, PhD Mentor: Hao Liang
- Faculty Mentor: Dr. Xinjie Lan
Flood
FlowState | State-of-the-Art Data-Driven Streamflow and Flood Forecasting for the Continental United States
This project develops a national-scale flood forecasting framework that predicts hourly streamflow 24 hours ahead during flood events across the Continental United States using five machine learning models. The models are trained on statistically defined flood events identified through Log-Pearson Type III (LP3) flood frequency analysis, following the USGS Bulletin 17C methodology. The framework integrates hourly meteorological data and static basin characteristics from the CAMELSH 2025 dataset, which covers 9,008 basins across the Continental United States. The models are evaluated using a variety of performance metrics, including Kling-Gupta Efficiency (KGE) and Mean Absolute Error (MAE), to assess predictive accuracy. The project also examines how each model handles diverse hydrological conditions across the U.S., with a focus on regional stratification and advanced feature engineering techniques to improve forecasting accuracy.
Student Team Members
- Gavin Daves
- Weizhe Mao
- Rahul Prakash
- Meaghan Ramlakhan
- Ricardo Rivera
- Pinar Targil
Sponsor and Mentors
- Sponsor Mentor: Dr. James Doss-Gollin
- D2K Fellow, PhD Mentor: Peikun Guo
- Faculty Mentor: Dr. Xinjie Lan
FMC_Safety
RAG Safe | TechnipFMC Safety: Historical Analysis of Safety Incidents
TechnipFMC has over a decade of global safety reports but analyzing these dense, unstructured texts is a massive challenge. We developed a two-layer AI pipeline that extracts key facts from these reports and organizes them into a searchable network of over 100,000 safety connections. The result is a powerful reasoning engine that can trace complex causal chains, helping the organization learn from the past to ensure a safer future.
Student Team Members
- Chloe Lim
- Keshav Shah
- Joel Villarino
- Isaiah Gonzalez
- Stanley Nwosu
- Jay Srivastava
Sponsor and Mentors
- Sponsor Mentor: Partha Dutta
- D2K Fellow, PhD Mentor: Johaun Hatchett
- Faculty Mentor: Chad Shaw
FMC Video
Hawkeye | Scene Understanding and Semantic Retrieval for Industrial Safety
This is an industrial AI video analysis system built for TechnipFMC. The project develops a vision-language pipeline for object detection and automated captioning on industrial video footage, generating natural language descriptions of video segments. To allow querying for videos using free-form keywords, an embedding-based semantic search was implemented to a web interface. It includes embedding-based semantic search so safety analysts can query video content using free-form text rather than exact keywords. A full-stack web application exposes the retrieval system, and a benchmarking suite evaluates model safety awareness against the iSafetyBench dataset.
Student Team Members
- Rafael Tinajero-Ayala
- Kunyang Li
- Richard Xu
- Marcos Miranda
- Laura Chirila
- Alan Johnson
Sponsor and Mentors
- Sponsor Mentor: Partha Dutta
- D2K Fellow, PhD Mentor: Nhi Le
- Faculty Mentor: Dr. Arko Barman
HFD
Delay Detectives | Estimating the Effect of Railroad Grade Crossings on Fire Department Response Times
Emergency response times are critical determinants of outcomes in both medical emergencies and fire suppression incidents. Our project estimates the causal impact of railroad grade crossings on Houston Fire Department response times in order to inform future policy and infrastructure decisions for problematic grade crossings.
Student Team Members
- Ananya Rao
- TJ Li
- Kelvin Phung
- Gia Kim
- Savan Patel
- Arkin Si
Sponsor and Mentors
- Sponsor Mentor: Leonard Chan, Michael Marino, Eric Flores
- D2K Fellow, PhD Mentor: Jose Palacio
- Faculty Mentor: Dr. Chad Shaw
ImpossibleArchieve
Team Impossible | Mapping the Impossible
We are working with the Center for the Impossible to make ~2,000 digitized letters to Whitley Strieber safer and easier to study at scale. We’re building a pipeline that cleans and chunks the text, anonymizes sensitive information (names, phone numbers, and emails), and then extracts subject–predicate–object “triples” that can be organized into a knowledge graph. The knowledge graph supports document-level tagging and a queryable database so researchers can search for patterns, like recurring beings, settings, time-of-day effects, or shared themes across letters, instead of relying on keyword search through long narratives.
Student Team Members
- JP Babin
- Grey Beaubien
- Mac Tucker
- Eugenia Nwogu
- Auggie Schwarz
Sponsor and Mentors
- Sponsor Mentor: Karin Austin and Amanda Focke
- D2K Fellow, PhD Mentor: Songyuan Sui
- Faculty Mentor: Dr. Shaw
Kinder
Research in Session | Automated Detection of Research Use in School Board Meetings
School board meetings across Houston area districts contain hundreds of hours of recordings where research and evidence are referenced, but identifying these moments manually is infeasible at scale. Our team built an end to end pipeline that scrapes board meeting videos, transcribes them using Parakeet ASR, and classifies 2 minute audio chunks as research mentions or not using a binary classifier trained on hand labeled transcripts. The system delivers flagged, timestamped transcripts to HERC (the Houston Education Research Consortium), allowing their team to track research use across 8 districts without sitting through thousands of hours of footage.
Student Team Members
- Melody Dao
- Annabelle Du
- Nilda Jarero
- Mirfat Maani
- Grant Thompson
Sponsor and Mentors
- Sponsor Mentor: Erin Baumgartner
- D2K Fellow, PhD Mentor: Ali Azizpour
- Faculty Mentor: Arko Barman
NASA
NASA Reentry | Predicting Spacecraft Reentry Conditions with Machine Learning
In this project, we aim to improve the prediction of heat flux during spacecraft re-entry. Current simulations (CFDs) are accurate but slow and expensive. We use machine learning trained on NASA data to quickly estimate heat flux across conditions. This enables faster design testing while maintaining high accuracy and safety.
Student Team Members
- Arush Adabala
- Ethan Hsu
- James Foxworth
- Kyle Zeng
- Livia Cordeiro
- Todd Hao
Sponsor and Mentors
- Sponsor Mentor: Andrew J. Hyatt
- D2K Fellow, PhD Mentor: Vishesh Kumar
- Faculty Mentor: Dr. Xinjie Lan
NatGEO
Team Chinchilla | Andean Rock Outcrop Classification and Chinchilla Occupancy Modeling
Industrial activity in the Andes is placing increasing pressure on the limited habitat of the short tailed chinchilla. By developing an algorithm to classify rock outcrops by geomorphology and identifying structural complexity metrics that most strongly support chinchilla occupancy, we aim to advance conservation efforts.
Student Team Members
- Jonathan Woo
- Allen Yuan
- Caroline Huynh
- J. Dante Maurice
- Jing Jing Chen
- Nomin Ganzorig
Sponsor and Mentors
- Sponsor Mentor: Alejandro Pietrek, Sofia Ocaranza Di Battista, Emilce Bustos, Andres Talamo, Julián Hernández
- D2K Fellow, PhD Mentor: Aditya Mandal
- Faculty Mentor: Arko Barman
OTT
Claim to Fame | A Data-Driven Approach to Patent Reference Classification
In preparing a patent application, determining whether cited references are material to patentability or simply serve as background information is currently a very time and resource intensive manual task. This project presents an automated framework for classifying patent references as material or background. Using transformer-based natural language processing techniques, a ranking is produced to sort the references from most material to background to enable users to quickly sift through a list of references. A confidence score is also calculated to allow the user to manually examine only the references that fall in the “grey area”.
Student Team Members
- Evan Brown
- Greyson Elyaderani
- Lucas He
- Saara Orav
- Vinay Joshi
- Kyle Sanderfer
Sponsor and Mentors
- Sponsor Mentor: Rice Office of Technology Transfer
- D2K Fellow, PhD Mentor: Janmajay Singh
- Faculty Mentor: Dr. Arko Barman
ROARS
ROARs | Rice Outcomes Assessment Reporting
The goal of this project is to automate the Rice Outcome Assessment Reporting process by the Office of Institutional Effectiveness. ROARs are vital to the success and development of academic programs at Rice, but are manual, time-consuming, and difficult to scale. Our solution is to create an AI tool to automate the review process and enable faster, more consistent program evaluation.
Student Team Members
- Jack Ma
- Samantha Ogundare
- Sai Raja
- Matthew Solomon
- Peyton Stevenson
- Zach Wilson
Sponsor and Mentors
- Sponsor Mentor: Diane Waryas Hughey
- D2K Fellow, PhD Mentor: Yehya Farhat
- Faculty Mentor: Chad Shaw
Smithsonian
Palyno-Minds | Automated Detection of Palynomorphs in Digital Microscopy
The Smithsonian stewards tens of thousands of microscopy slides containing fossil palynomorphs (organic microfossils including pollen, spores, fungi, and algae) yet the sheer volume of specimens makes manual analysis infeasible. In this project, we design a scalable end-to-end pipeline to automate the detection and localization of palynomorphs within digitized microscopy slides. We develop an efficient pre-processing workflow to tile and compress high-resolution and multi-focal images, and leverage a transformer-based model, RF-DETR, to enable accurate detection in dense and morphologically diverse samples. Our approach enables fast whole-slide annotation, transforming a previously time-consuming manual task into an efficient, automated workflow for palynological research.
Student Team Members
- Abbas Shaikh
- Teon Golden
- Aditya Viswanathan
- Praise Mayor
- Patrick Ainlay-Vazquez
- Eric Zhang
Sponsor and Mentors
- Sponsor Mentor: Dr. Ingrid Romero, Dr. Scott Wing
- D2K Fellow, PhD Mentor: Tony Yu
- Faculty Mentor: Dr. Arko Barman
TicketMastery
TicketMastery | An Interactive Dashboard for Highlighting Chronic Building Issues on Rice Campus Through Maintenance Requests
Rice Facilities & Capital Planning (FC&P) manages the Rice maintenance request system, where students, faculty, and staff submit work orders when they notice issues in campus buildings. FC&P has collected data describing these so-called “tickets,” which represent maintenance tasks to be completed around campus by service technicians. We built a dashboard to help FC&P quantify and identify work orders that were not successfully resolved, and to identify the factors contributing to the mitigation of chronic facility defects on campus.
Student Team Members
- Beck Edwards
- Elijah Afrifa
- Ethan Carr
- Lucca Ferraz
- Maximus Cenni
- Jimmy Li
Sponsor and Mentors
- Sponsor Mentor: Terie McClintock (Facilities & Capital Planning, Rice University)
- D2K Fellow, PhD Mentor: Caleb Skinner
- Faculty Mentor: Dr. Xinjie Lan
Alzheimer’s Prediction
N.E.U.R.O.N (Neuroimaging and Event-based Unified Risk Outcomes Network) | Forecasting the Cognitive Clock: Survival Modeling for Alzheimer's Progression
Alzheimer's disease develops silently, and often decades before a diagnosis. We built a multimodal, multi-state survival analysis system that predicts how long a patient with Mild Cognitive Impairment will remain stable before converting to Alzheimer's Disease. The result is a tool that could help clinicians intervene earlier, stratify patients more precisely, and accelerate clinical trial enrollment.
Student Team Members
- Savannah Nix
- Evie Roth
- Eliza Iqbal
- Nathon Chavez
- Omar Dajani
- Fabrizio Pacheco
- Shichen Tang
Sponsor and Mentors
- Sponsor Mentor: Cindy Zhang
- D2K Fellow, PhD Mentor: Antonio Mendoza Gonzales
- Faculty Mentor: Xinjie Lan