AP / ENGINEERING OBSERVATORYSOFTWARE · DATA · INTELLIGENCEChapel Hill, NC

Ideas, connected. Systems, built.

From ideas to working systems.

I’m Ayug, a Computer Science and Statistics student at UNC–Chapel Hill. I build full-stack applications, AI-powered tools, and backend systems that turn complicated problems into useful software.

Explore projectsThree complete applications, examined up close

01—06Messaging. Applications. Data.
Follow the shared engineering.

The project observatory06 PROJECTS · ONE PERSPECTIVE

AI & agentsTAB

A HackNC 2026 expense agent that turns iMessage conversations and receipt photos into trip ledgers, dispute workflows, and Solana devnet settlements.

TypeScript / Hono / SQLite / Photon / Gemini / Solana devnet

Open complete case study

Choose a project coordinate

Making complex systems feel simple.Open the case files

01Engineering case files

Open the case files.

The problem, the decisions, and the system that came out of it. Three builds, examined up close.

AP—01ENGINEERING RECORD

TAB / SYSTEM OVERVIEWCONCEPT DIAGRAM

iMessage / illustrative input“I paid for gas. Split with everyone.”

PhotonGemini intentTyped action

Deterministic ledgerInteger cents · SQLite · disputes

Solana devnetTest-token settlement

IMESSAGE → AI INTENT → LEDGER → DEVNETFIG. 01

01/ 03AI agents / Financial systems

TAB

A message in. A balanced ledger out.

A HackNC 2026 expense agent that turns iMessage conversations and receipt photos into trip ledgers, dispute workflows, and Solana devnet settlements.

Built withTypeScriptNode.jsHonoPhoton SpectrumGeminiSQLiteSolana

Public source · team projectHackNC 2026 / devnet

Read case file

AP—02ENGINEERING RECORD

GRADESENSE / REQUEST FLOWCONCEPT DIAGRAM

01 / REQUESTAppeal submitted

FastAPI + SQLModel

02 / INTELLIGENCEAI preview + recommendation

Role-based access

03 / REVIEWStaff decision → final score

STRUCTURED DATA. CLEAR RESPONSIBILITIES.FIG. 02

02/ 03Full-stack / Applied AI

GradeSense

Give every appeal a structured path.

An end-to-end exam and regrading feature inside LearnWithAI: graded PDFs, question rubrics, AI-assisted appeals, instructor decisions, and final-score tracking.

Built withAngularTypeScriptFastAPIPythonPostgreSQLSQLModelAzure OpenAI

Private source · team projectComplete LearnWithAI feature

Read case file

AP—03ENGINEERING RECORD

EMPOWHER / CONNECTED TOOLSCONCEPT DIAGRAM

FOUR ENTRY POINTS. ONE PLATFORM.FIG. 03

03/ 03Hackathon winner / Applied AI

1st place · Fidelity FidHacks 2025

EmpowHER

Four tools. One next step forward.

An AI-powered career-development platform connecting interview practice, resume analysis, learning resources, and event discovery.

Built withHTML/CSSJavaScriptHTTP APIsFileReaderServor

Public source · team projectFidHacks 2025 / winner

Read case file

Beyond selected workPorchBackend / iOS community platform Materials ResearchPython / Computational science

AEngineering archive

Smaller builds. Useful experiments.

Public repositories beyond the featured work, with their implementation scope kept visible.

A—01

Data engineering experiment

Bet-Tracker

A season-data collection script and a CLI that computes player-stat distributions using a normal CDF. A useful early example of a collection → CSV → analysis pipeline.

Python / Pandas / SciPy / CSV

Exploratory statistical code with local-path assumptions. No validated prediction accuracy or wagering performance is claimed.

A—02

Early full-stack scaffold

BETAPP

A frontend/backend starter with FastAPI health endpoints and a React application structure. Included as an exploratory repository, separate from the featured applications.

React / FastAPI / Axios

The repository marks prediction, authentication, persistence, and ML integration as future work.

Browse the GitHub profile

02Engineering log

A record of building.

From industry tooling to scientific research. Taking responsibility for the systems behind the interface.

Jun — Aug 2026Industry

Fidelity Investments

Full-Stack Software Engineering Intern

Making internal AI work easier to find, explore, and publish.

Architected an Angular/TypeScript developer hub consolidating over 40 versioned AI prototypes.
Integrated application and version-resolution APIs, taking discovery from four steps to two.
Implemented authenticated CRUD and version-management workflows for self-service publishing.

40+AI prototypes unified

<30sDiscovery, from 3–5 min

Prototype discovery, consolidatedARCHITECTURE / ILLUSTRATION

FRAGMENTED DISCOVERYBefore
4 discovery steps · 3–5 minutes
40+ VERSIONED PROTOTYPESDeveloper hub
Angular + TypeScript · searchable catalog
INTEGRATIONAPI resolution
Application + version-resolution APIs
SELF-SERVICE DISCOVERYAfter
2 steps · under 30 seconds

Authenticated CRUD → version management → self-service publishing

Original before/after illustration using supplied résumé metrics. No internal company interface or source code is reproduced.

Sep 2026 — PresentCommunity

App Team Carolina

Backend Developer · Porch

The server-side foundation for a more connected neighborhood.

Developed backend APIs and data models for a community-focused iOS application.
Implemented server-side application workflows.
Collaborated with iOS, product, and design teammates.

The server-side foundation of a neighborhood appARCHITECTURE / ILLUSTRATION

DOMAINCommunity data
Events · resources · neighborhood information
ENGINEERING SCOPEBackend
APIs · data models · server workflows
CLIENT INTEGRATIONiOS experience
Community information in the mobile application

Backend ↔ iOS, product, and design collaboration

Conceptual scope illustration based on supplied experience. No unverified backend framework or public repository is claimed.

Jun 2023 — Jun 2024Research

Fayetteville State University

Computational Materials Research Assistant

Turning simulation outputs into scientific understanding.

Built Python and NumPy/Pandas processing pipelines for simulation outputs.
Automated five scientific analysis workflows.
Studied Sc₂C MXenes using Quantum ESPRESSO outputs, including band structures, density of states, Fermi-level, and charge-density analysis.

05Analysis workflows automated

From simulation output to interpretable dataARCHITECTURE / ILLUSTRATION

COMPUTATIONAL INPUTQuantum ESPRESSO
Simulation outputs for Sc₂C MXenes
FIVE ANALYSIS WORKFLOWSPython pipelines
NumPy + Pandas · automated processing
SCIENTIFIC OUTPUTAnalysis
Bands · density of states · Fermi level · charge density

Original processing-pipeline illustration. It presents the analysis workflow without inventing material-property results or numerical plots.

03About / Technical profile

The person behind the systems.

I like the moment when a complicated problem starts to make sense.

That’s what connects my work across full-stack applications, AI agents, backend engineering, and research. I’m interested in how the pieces fit together—and how to make the result useful to the person on the other side of the screen.

Currently studying

University of North Carolina
at Chapel Hill

B.S. in Computer Science and Statistics
Double Major · Expected May 2028

35.9049° N / 79.0469° W

The engineering toolkit

Languages

Java, Python, JavaScript, TypeScript, HTML/CSS

Frameworks

Angular, React, FastAPI, Hono, Node.js, Spring Boot, Bun, SQLModel, SQLAlchemy

Data

PostgreSQL, SQLite, MySQL, Pandas, NumPy, Matplotlib

AI & tools

Azure OpenAI, Gemini API, Photon Spectrum, Solana devnet, Docker, Git, Jupyter, Quantum ESPRESSO

Testing

Playwright, Pytest, JUnit, Mockito

+Different tools. Connected thinking.

From ideas to working systems.

Open the case files.

TAB

GradeSense

EmpowHER

Smaller builds. Useful experiments.

Bet-Tracker

BETAPP

A record of building.

Fidelity Investments

App Team Carolina

Fayetteville State University

The person behind the systems.

University of North Carolinaat Chapel Hill

Languages

Frameworks

Data

AI & tools

Testing

University of North Carolina
at Chapel Hill