AVA – AI Virtual Assistant for Workflows

AVA (AI Virtual Assistant) started from a simple frustration: too much repetitive “click work.”
Updating Excel sheets, copying charts into PowerPoint, exporting dashboards — all tasks that drain time without adding much value.

I wanted something that could understand intent, coordinate tools, and deliver results the way a real assistant would.

Why I Built It

In both academic and professional work, I kept seeing the same bottleneck:
smart people wasting hours manipulating files instead of thinking strategically.

My goal was to design an assistant that users can speak to naturally:

“Update the quarterly sales Excel file and prepare a 4-slide summary.”

AVA interprets the request → calls the right modules → and returns finished output.

Technical Architecture

The system uses a modular skills design:

  • Excel Skill: read/write, pivot tables, cleaning
  • PowerPoint Skill: generate slides, insert charts, export PDFs
  • Data Skill: run SQL queries, filter datasets, export CSV
  • Dashboard Skill: trigger refreshes, capture snapshots

A central AI Orchestrator maps user intent to skills:

  1. Intent parsing
  2. Task sequencing
  3. Execution, logging, confirmation
  4. Human approval when needed

What AVA Can Do Now

  • Produce slide decks from spreadsheets
  • Refresh data and export summaries
  • Clean tables and generate insights
  • Automate repetitive reporting sequences

What I Learned

Building AVA taught me that automation isn’t about replacing humans —
it’s about eliminating tedious work so people can focus on thinking.

This project continues to evolve, but it represents my broader interest in
AI systems that create real, measurable productivity.

Stay updated

Occasional updates on projects, lessons learned, and experiments in AI and data.