I build web appsfrom interface to API.

We wanted transactions, budgets, reports, and spending forecasts to feel like one app instead of separate features.

I worked on the frontend in a six-person capstone team, building the React screens and connecting them to the team's API and AI services.

I built the React interface for entering transactions, checking budgets, reading reports, and viewing predictions, then connected those screens to the team's backend.

The team delivered a working capstone with eight spending categories and a classifier trained on 45,000 records that reached 98.79% holdout accuracy.

Comparing dozens of product descriptions is slow, especially when users only know the result they want rather than a product name.

I built the JavaScript frontend, FastAPI backend, Supabase catalogue, admin page, and content-based recommendation flow.

I connected the chat interface to a FastAPI service that compares the user's request with the catalogue using TF-IDF and cosine similarity.

The finished app includes a 70-product catalogue, product management, and ranked top-three recommendations.

Smart-contract interactions are difficult to understand when users only see wallet prompts and transaction data.

I worked with React, Solidity, Hardhat, ethers.js, and the Mantle network to connect the interface with the smart contracts.

I built a React interface that turns contract actions into familiar steps for creating, exploring, and managing royalty tokens.

The result is a working prototype on Mantle. It demonstrates the idea, but it is not a live investment product.

The screening needed clear rules and saved results so each outcome could be traced back to the symptoms selected.

I built the FastAPI and MySQL backend, the rule-processing flow, stored diagnosis sessions, and a small administration area.

I represented the symptom relationships as forward-chaining rules, exposed them through FastAPI, and stored each session in MySQL.

The prototype can run the same rules consistently, save the result, and be set up locally with Docker.

The original client only worked comfortably from a terminal, and repeated requests wasted sessions and API limits.

I moved the existing Python workflow into FastAPI and built the browser interface, multi-account switching, and token caching.

I built a small JavaScript SPA around FastAPI endpoints and used SQLAlchemy with PostgreSQL caching, with SQLite available for local development.

The app is deployed on Vercel and supports multiple accounts, stateless requests, database caching, and local setup.

Customers needed one place to check the food, prices, opening information, location, and ordering options.

I built the Next.js frontend, organized the menu data, and added category filters, cart interactions, Maps, and direct ordering links.

I split the interface into reusable React components for the landing page, menu, cart, ordering links, and location, then adapted the layout for desktop and mobile.

The deployed site includes a searchable menu, responsive cart, business details, Maps integration, and local-business metadata.