Pulse AI: Predictive Forecasting Engine
Category:
AI-Augmented Data Engineering / Predictive Analytics
Client:
Self-Initiated (Hope Tech R&D)
The Problem
Many small businesses and individuals have historical data but no easy way to project future performance without hiring a data scientist. I noticed a gap between "simple spreadsheets" and "complex enterprise software." I wanted to see if I could build a professional-grade forecasting tool that was simple enough for anyone to use, while also pushing my own technical boundaries by working with unfamiliar libraries like Facebook Prophet.
The Solution
I developed Pulse AI, a web-based analytics engine that automates time-series forecasting.
Vibe-Coding Methodology: I acted as the lead architect, managing a "digital team" of AI models (Gemini, Claude, and DeepSeek) to write the foundation, fix complex UI bugs, and optimize data processing.
Predictive Power: Integrated the Prophet algorithm to allow users to generate daily, weekly, or monthly projections instantly.
Data Integrity: Built-in anomaly detection to flag unusual data spikes and a professional PDF reporting system for executive-level summaries.
The Result
The project resulted in a fully functional, cloud-deployed application that simplifies complex data science for everyday users.
Productivity Boost: By leveraging AI collaboration, I reduced the development time by an estimated 70% while maintaining high code quality.
Engineering Honesty: After testing the AI Chat feature, I made the strategic decision to disable it due to reliability concerns, prioritizing Data Accuracy over "flashy" features.
Technical Success: Successfully integrated a modern tech stack (Streamlit, Supabase, Google Analytics) into a cohesive, user-friendly product.
