Outflow

Built an AI-powered personal finance platform using Flask and PostgreSQL, featuring a conversational assistant capable of adding, updating, deleting expenses and answering spending queries through natural language.

Designed an AI pipeline using Groq Llama 3 with intent routing, structured JSON responses, voice-to-text transcription via Whisper, and human confirmation before database updates.

Implemented production-oriented AI guardrails including Pydantic schema validation, rate limiting, intent validation, business rule enforcement, and database isolation to ensure reliable and safe LLM interactions.

Engineered a responsive server-rendered frontend with Jinja2, Vanilla JavaScript, AJAX-powered interactions, interactive analytics, custom CSS, dark mode, and an instant Demo Mode with isolated user sessions.

Implemented production focused backend architecture using raw parameterized SQL, Flask Limiter, CSRF protection, secure session based authentication, Alembic database migrations and strict separation between AI reasoning and database execution.

Containerized the application with Docker and deployed it to Microsoft Azure App Service using Gunicorn, Azure Container Registry and Neon PostgreSQL.