A.B. Duke Scholar · ML Engineer

Joshua Chilukuri

Excited about the vision and language modalities of AI and their applications in improving accessibility and safety. Interested in the study of AI interpretability, transparency, and ethics.

Learn More

Snapshot of my life

Fig. 02 — About
Joshua Chilukuri
Joshua Chilukuri — Portrait

Joshua Chilukuri

Hey there! My name is Joshua Chilukuri and I'm an incoming Angier B. Duke Scholar at Duke University, pursuing my interests in computer systems, AI, and public policy. Today, I'm most interested in multi-modal AI as it relates to improving human lives by improving safety and reducing accessibility barriers. My research spans augmented reality, 3D content generation, computer vision/perception, and AI ethics.

01

Finance Secretary

UNC Association of Student Governments

02

XR + AI Researcher

Duke University — I3T Laboratory

03

Co-Founder

Hiant Development

04

Prev. Data Analyst Intern

Lenovo Corporation - Supply Chain Team

I am currently based in San Francisco where I will be working at a startup. Feel free to reach out if you're nearby!

View Projects
Fig. 03 — Projects
Recognition
Year

2026

3 Projects
Distill
1st Place @ GitHub Copilot Hackathon NYC

Distill

Distill turns dense research papers into concise, visual summaries. Upload a PDF and it extracts, chunks, and synthesizes the key ideas in parallel into structured Markdown and a single self-contained, embeddable HTML visualization. Awarded 1st Place & $25K in Azure Credits at MSFT GitHub Copilot Hackathon NYC

AINLPFRONTENDBACKENDGEN AI
Local402
MSFT × Coinbase × Tavily Build Day

Local402

A native macOS assistant that runs a local LLM over your private documents and decides when to pay for live web search, settling each query in USDC via x402 micropayments on Base. Your documents stay on-device and only deliberate search strings ever leave.

AINLPFRONTENDBACKENDDBGEN AI
Kirin
Personal

Kirin

Kirin is a lightweight, transparent AI assistant overlay for Windows OS that enables you to quickly query AI with screen context. It preserves privacy by enabling you to specify the screenshot window.

AICVHARDWAREDS/MLBACKENDGEN AI

2025

9 Projects
AVISION (In-Progress)
Personal

AVISION (In-Progress)

Currently building touch-activated AI glasses interface, designed to access custom APIs for blind navigation, data extraction, and object memorization.

AICVHARDWAREDS/MLBACKENDGEN AI
North Carolina DECA App
10K+ Members Served

North Carolina DECA App

In partnership with NC DECA (10K+ members statewide), the DECA App serves as a communication resource featuring authentication, push notifications, resource links, and member filtering, to connect state officers DECA members across NC.

AICVHARDWAREDS/MLBACKENDGEN AI
Indoor-RFDETR
HPC-Trained Model

Indoor-RFDETR

Trained using a high performance computing cluster, Indoor-RFDETR is a fine-tuned variant of the Roboflow Detection Transformer architecture for identifying key indoor objects (staircases, doorways, furniture, etc) to enhance blind navigation when implemented in smart glasses designs.

AICVDS/ML
Voice-to-3D Object Generation Engine
Best Paper @ UNAI Workshop (IEEE ISMAR)

Voice-to-3D Object Generation Engine

Research project with Duke University's I^3T Lab -> Building methods for automated, real-time content generation in Mixed Reality (XR) environments. Awarded Best Paper at Augmented Interaction Workshop, IEEE ISMAR.

NLPAR/VRGEN AIRESEARCH
SightLine
Intel Global AI Award

SightLine

AI-Glasses for the visually impaired. Leverages monocular metric depth estimation, YOLOv8 object detection, and image streaming to produce contextual system prompts for a cloud-hosted LLM, allowing specific, depth-informed navigation capabilities.

CVNLPAR/VRiOSHARDWAREGEN AI
North Carolina TSA App
800K Requests · 2K Users

North Carolina TSA App

Serving 800k backend server requests, and downloaded by ~2k conference attendees, the North Carolina TSA App managed event data, notifications, and provided information tools for one of North Carolina's largest career technical service organizations (9k+ members).

iOSFRONTENDBACKENDANDROID
Predicting U.S Food Insecurity: An Empirical Assessment of ML Approaches and Sociodemographic Predictors
Published

Predicting U.S Food Insecurity: An Empirical Assessment of ML Approaches and Sociodemographic Predictors

Independent research, evaluating XGBoost, Random Forest, Neural Networks, and Google TabNet as predictive models for classifying food deserts using publicly available census data.

AIDS/MLRESEARCH
Ketchup
Ready, Set, App! Challenge

Ketchup

A mobile app designed for integration in elementary/middle schools, leveraging LLM's, PDF document uploads, and forum functionalities, to create a modern tool for teachers to support students catching up on absent work.

AINLPiOSFRONTENDBACKENDANDROID
ARES-M Weather Sensor
1st Place @ SMathHacks 2025

ARES-M Weather Sensor

ARES is a remote autonomous sensor suite, designed for deployment on Mars, monitoring temperature, humidity, seismic activity, UV/IR, and soil moisture in real time.

HARDWAREDS/MLFRONTEND

Press · Recognition

News & Features

Intel Global AI Award Announcement01 / 04

Intel Global AI Award Announcement

Featured on Intel Artificial Intelligence Newsroom and Intel.com as a recipient of the 2025 Intel Global AI Impact award. Demonstrating a novel approach to metric-depth informed AI perception, applied to Smart Glasses to create a navigation interface for the visually impaired.

Read the feature
DECA International Grand Award (NCSSM Entrepreneurship)02 / 04

DECA International Grand Award (NCSSM Entrepreneurship)

Featured for representing NCSSM at the DECA International Career Development Conference (ICDC) 2025 in Orlando, FL. Awarded 2nd in Principles of Finance against hundreds of global finalists worldwide.

Early Careers at Lenovo (HS Intern Day in the Life)04 / 04

Early Careers at Lenovo (HS Intern Day in the Life)

One of two high school interns featured for Lenovo's 'Day in the Life' Intern marketing campaign on Lenovo Early Careers. Video briefly highlights my experience on Lenovo's Procurement Team.

Research · Writing

Publications & Articles

  1. 01

    Cognitive Attack Detection in Augmented Reality (CADAR): A Neuro-Symbolic Approach with Particle Filtering on Perception Graphs

    Chen R., Andreyev A., Xiu Y., Chilukuri J., Sen S., Imani M., Li B., Gorlatova M., Tan G., Lan T.

    Published in IEEE CVPR Findings 2026

  2. 02

    Say It, See It: A Systematic Evaluation on Speech-Based 3D Content Generation Methods in Augmented Reality

    Xiu, Y., Chilukuri, J., Sen, S., Gorlatova, M.

    Published in IEEE ISMAR UNAI Workshop 2025

  3. 03

    Predicting Low Food Access: An Empirical Assessment of Machine Learning Algorithms and Socio-Demographic Predictors

    Chilukuri, J.

    Preprint available on IEEE TechRxiv

  4. 04

    ARAM Back Band: An Entrepreneurial Business Plan and Proposal for a Hypothetical Revolutionary Medical Wearable in the Cramp Detection and Relief Market

    Chilukuri, J., Kela, S., Jha, A., Sah, A.

    Frontier (NCSSM Journal of Economics and Entrepreneurship)