Nexora

Chicago has over 600,000 freelancers and independent contractors who are underserved by tools built for salaried employees — they overpay taxes, miss deductions, and have no real cash flow visibility. That gap inspired Nexora. We built a full-stack platform that connects to real bank accounts via Plaid, automatically categorizes transactions against IRS Schedule C, computes quarterly tax estimates in real time, forecasts cash flow using custom ML models, and delivers personalized financial advice through a GPT-4o multi-agent system. The stack runs on FastAPI, PostgreSQL, React 19, and Docker, with MLflow tracking our training pipelines. The hardest challenges weren't technical — grounding AI advice in real transaction data without hallucination, handling the irregular income patterns freelancers live with, and earning the trust that comes with storing someone's banking credentials pushed us to build with security and accuracy as first-class concerns from day one. We came away with a working platform we're genuinely proud of, and a deep respect for the complexity hiding behind what looks like a simple budgeting app.

Built With

• alembic
• celery-databases:-postgresql
• jwt
• kubernetes
• languages:-python
• numpy-infrastructure:-docker
• openai-gpt-4o-ml-&-data:-mlflow
• pandas
• plaid
• react-19
• recharts
• redis-apis:-plaid
• scikit-learn
• sqlalchemy
• tailwindcss
• tanstack-query
• terraform-other:-vite
• typescript-frameworks:-fastapi
• zustand