{
"vibes": [
"plays chess",
"breaks CTFs",
"smashes TT",
"thinks about AGI too much",
"reads existential psychology for fun",
"runs on coffee with extra sugar ☕",
"turns napkin ideas into live URLs"
]
}I build AI systems that don't just predict — they reason, orchestrate, and self-improve. Multi-agent swarms that get better every cycle. Adversarial training loops that harden models while you sleep. RAG pipelines that actually work in production. IDE skills that make other developers' agents smarter.
The first platform that actively hardens models via biologically-grounded sleep cycles.
Wake → Dream → Nightmare → Compress → repeat
Production models silently degrade — adversarial perturbations collapse accuracy from 92% to 23%. NightmareNet fixes that with a cyclic training paradigm inspired by sleep-mediated memory consolidation. Each cycle produces models that accumulate robustness without catastrophic forgetting.
- 500+ tests passing | Dockerized | EU AI Act Article 15 compliance reporting
- +13.64% relative robustness WITHOUT clean-accuracy tradeoff
- 3 cycles: 58.4% TextFooler acc vs 23.1% baseline — with 35% fewer parameters
- Cyberpunk Next.js 16 dashboard (20 panels) + CLI (
nightmarenet train | evaluate | benchmark) - Apache 2.0 OSS core — runs on a 4GB VRAM laptop GPU
| Project | What it does | Link |
|---|---|---|
| NWTR | AI-powered real estate — Next.js 15, Three.js, R3F, GSAP, Spline, 42-component design system | Live |
| AtomicPulse | AI performance management — GPT-4o copilot, SMART goals, KPI prediction, semantic search | Live |
| PrepWise.AI | Azure OpenAI interview platform — resume parsing, adaptive questions, tone analysis, scored PDFs | Live |
| MediaOS | AI media buyer agent — audience research engine (6 providers via Bright Data), ad creative gen, landing page deploy | Live |
| Vibe Syndicate | 6-agent self-improving swarm — cross-model review, compound memory, 14 MCP tools | Live |
| XenoReach.AI | AI-native CRM — autonomous campaign agents that plan, execute & optimize marketing workflows | Live |
| Workforce Dashboard | Regional workforce intelligence — Indeed/LinkedIn/USAJobs ingestion, skills gap mapping, NL Q&A | Live |
Also: SaarthiAI (11-language farmer voice assistant, RAG) · AI Tool Discovery Agent (async GPT-4o pipeline + Teams) · GitHub AI Code Review Bot (RAG + Azure Functions)
Self-improving intelligence amplifier for AI coding agents. 18 behavioral skills (diagnose, prove-it, self-review, grill, context-builder, mcp-conductor, ai-debt-detector...) that make Cursor, Claude Code, Copilot, and Codex fundamentally better. Zero npm dependencies, TF-IDF routing, SkillOpt self-improvement loop.
npx skills@latest add Adit-Jain-srm/skill-forge14-tool MCP server that exposes a full multi-agent swarm workflow from your IDE. Submit tasks, trigger cross-model reviews, query compound memory, manage approvals — all without leaving Cursor.
{ "mcpServers": { "syndicate": { "command": "python", "args": ["syndicate-mcp/server.py"] } } }Merged PRs to external repos:
| Repository | PRs | Highlights |
|---|---|---|
| Him-an-shi/Aarini | 32 merged | SSE streaming, i18n, encryption, offline sync, accessibility, auth, full test suite |
| PRODHOSH/ossfolio | 11 merged | Profile discovery, keyboard shortcuts, OG tags, heatmap, type safety |
| SRV30/Faith-and-Fast | 5 merged | Security hardening, rate limiting, env validation, perf, error boundaries |
| souma9830/SnapPass-AI | 4 merged | Rate limiting, Redis graceful degradation, a11y, meta hooks |
| SAARTHII-AI/Saarthi-AI | 3 merged | Azure + Brightdata integration, frontend refactor |
| SRV30/samridhi-enterprises | 3 merged | Env validation, structured logging, error handling |
| akshay0611/havenly | 3 merged | Review form, share property, dynamic UI |
| arghya29/EcoSphere | 3 merged | Loading/error states, UI fixes |
| intuit/quickbooks-online-mcp-server | 2 merged | LinkedTxn feature, OAuth docs |
| SRV30/Box_Office_Inc-Movie_Sim | 2 merged | Feature contributions |
| fastify/fastify (36k+ stars) | 1 merged | Docs: fix incorrect defaults and code examples |
| firecrawl/firecrawl | 1 merged | Python SDK Pydantic fix |
| wshobson/agents | 1 merged | skill-forge-essentials plugin (3 novel behavioral skills) |
| kunovsky/paid-open-source-projects | 1 merged | tscircuit listing |
| souma9830/FixNearby | 1 merged | Password strength validation |
| arghya29/Github-User-Analyser | 1 merged | Expandable repo details |
74 merged PRs across 16 external repos — features, security, accessibility, testing, performance
The full story — experience, certs, education, and everything in between:




