A1
AI coding agents
Daily drivers for generation, review and refactoring — orchestrated, not just prompted.
SYSTEM ARCHITECT · AI ENGINEER · INDIE HACKER
DINUDSI
I build what others imagine.
›
13+ years of production systems, fused with AI-native agent workflows — MVPs in days, security-first, shipped fast.
13+
Years shipping
37
Public repos
659
GH followers
80%
Cost cut
AI-native
by design
I'm Dinudsi — a System Architect based in Udaipur, India, shipping software since 2009. I design enterprise-grade backends and secure APIs, then multiply the output with AI agent workflows so teams get MVPs in days, not months — at ~80% lower cost.
The core is Spring Boot, Node.js, PostgreSQL. The accelerator is Claude Code, Codex, OpenCode, Copilot CLI. The glue is custom MCP servers, agent memory systems, and security-first code review on every merge.
Outside client work I build in public — Docly, Commitify, and currently strangers.lol.
Agent-first
toolkit
Everything I reach for to design, build, review and ship production systems — from coding agents to vector stores.
A2
Agent infrastructure
B1
Security & review
B2
Backend
C1
Data & AI
C2
DevOps & cloud
D1
Languages
D2
Frontend & OSS
Selected
builds
Indie products, agent tooling and production systems — each one built to survive contact with real users.
strangers.lol
A new real-time social experiment — pitched from Next.js to Vite + React + Fastify when persistent WebSockets demanded it. Day-by-day build logs on X.
Visit siteDocly
Browser markdown editor with live preview, PDF export, syntax highlighting and unlimited saves. No paywall — Google sign-in only if you want the cloud.
Open DoclyBrainSync
Cross-agent memory layer. Lets AI coding agents retain knowledge and share context across sessions via vector embeddings.
Claude Security Marketplace
Marketplace of advanced domain-specific skills, hooks, agents and commands — security-focused tooling for AI workflows.
GitHubCommitify
VS Code extension that auto-commits your changes every 5 minutes with Copilot-generated commit messages. Pinned on GitHub.
GitHubBrainByte
A transformer language model written from scratch in PyTorch — trained for structured tool-calling via next-token prediction.
collab-cli
Universal collaboration protocol + CLI for multi-agent LLM teams — Claude Code, Codex and friends working as one unit.
GitHubAI CCTV System
Real-time plate recognition, violence detection and threat analysis — YOLOv5 + vision-language models over live video.
Operating
system
How one architect + a swarm of agents delivers production software at 10× speed.
Multi-agent orchestration
Role-based agents for generation, review, testing and deployment — coordinated through custom orchestrators.
Custom MCP servers
Tools and integrations that extend what agents can touch — files, APIs, internal systems.
Agent memory systems
Persistent context (BrainSync) so agents remember decisions across sessions and teammates.
Security-first review
Every PR: AI-assisted vulnerability detection, secret scanning and threat modeling before merge.
I don't just write code — I orchestrate AI systems that write, review, and ship code.
01
Architecture first —
clarity before code
02
AI augmented — agents as
multipliers, not replacements
03
Security by default —
rigorous, secure by design
04
Ship fast — MVPs in days,
scaled in weeks
Let's build
something loud
Open to new MVPs, agent tooling and out-of-the-box ideas. Pick a channel — I reply fast.