Co-built · architected the backend
AAVA Code
AI Coding Plugin for VS Code · Ascendion
- 3K+
- daily users
- 5+
- client environments
- 150+
- skills · 40+ tools · ~60 commands
Problem
AAVA Code began as a single-agent prototype — fine for demos, but it couldn't scale to the breadth of skills, tools, and commands real client environments demanded, and a single agent became a bottleneck for complex coding tasks.
Approach
I co-built the backend for AAVA Code and architected its re-design from that single-agent prototype into a multi-agent orchestration system: a Main-Agent + sub-agent topology via crewAI flows, exposing 150+ skills, 40+ tools, and ~60 commands. I also contributed to the VS Code plugin frontend so the agent experience surfaced cleanly in-editor.
Impact
- Adopted by 3K+ users daily across 5+ client environments.
- Multi-agent orchestration replaced the single-agent ceiling.
- 150+ skills · 40+ tools · ~60 commands in a governed plugin.
Contribution: Co-built the plugin; I majorly contributed to and architected the backend, and also contributed to the VS Code frontend.
// constraints
The single-agent prototype already had clients in 5+ environments — the re-architecture had to be backward-compatible with existing VS Code extension contracts while expanding to 150+ skills without a flag day migration.
// tradeoffs
Used crewAI flows for the Main-Agent → sub-agent topology instead of a custom orchestrator: faster to ship and battle-tested for tool-calling, but constrained by crewAI's execution model. Traded orchestration flexibility for delivery speed — the right call under the timeline.