harsha = {
"age" : 18,
"location" : "Andhra Pradesh, India 🇮🇳",
"approach" : "vibe coding — AI as creative co-pilot",
"tools" : ["Claude", "Cursor", "Emergent AI"],
"stack" : ["Next.js", "TypeScript", "Supabase", "Python"],
"building" : "CreatorStack 🎬 — LinkedIn for video creators",
"researching": "DENA 🧠 — Distributed Expert Neural Architecture",
"competing" : "The Anvil Hackathon @ Scaler SST, Bengaluru 🏆",
"open_to" : ["hackathon teams", "collabs", "good problems"],
}| Project | Description | Stack | Status |
|---|---|---|---|
| 🧠 DENA | Distributed Expert Neural Architecture — modular AI research with dynamic routing, aggregation & memory | Python · Research |  |
| 🎬 CreatorStack | Professional network for video editors & motion designers — LinkedIn × Behance × Fiverr for the video economy | Next.js · Supabase · Cloudflare Stream |  |
| 🪄 PostCraft Pro | AI LinkedIn post generator with competitor analysis & scoring | Next.js · Supabase · Sarvam · TypeScript |  |
| 🖥️ Swabox | AI-integrated terminal with plugin system | Python · Docker · Rich |  |
DENA is an independent research project exploring how modular AI systems can route tasks to specialized expert subsystems instead of relying on a single dense model.
The architecture studies:
- Task decomposition — breaking queries into routable subtasks
- Dynamic expert routing — directing tasks to the right specialist
- Weighted aggregation — combining expert outputs intelligently
- Verification + memory — ensuring quality and learning over time
Status: DENA v0 is a preliminary prototype using rule-based experts and controlled benchmarks. Results show measurable cost-aware routing but do not claim superiority over production MoE systems.
📄 Paper: DENA: A System-Level Architecture for Distributed Expert AI Systems
🔗 DOI: 10.5281/zenodo.19871653
📦 Repo: harsha-sai-q/DENA
+ researching → DENA v0.3 — distributed expert neural architecture (published on Zenodo)
+ building → CreatorStack — professional network for video creators
+ competing → The Anvil Hackathon @ Scaler School of Technology, Bengaluru
+ learning → full-stack fundamentals · Git · prompt engineering
+ open → hackathon teams · collabs · interesting problems
building different · shipping fast · always learning