Engineering category-defining developer tools, zero-latency AST invariant radars (<50ms), and parallel reapers that reclaim gigabytes of disk debris and kill zombie dev servers.




//Contextual Codebase Companions & Full-Stack Scaffolding
Self-directing agentic loops navigating deeply nested architectures, generating verified full-stack applications, and conducting multi-source research from single prompts.
//Deep Graph Topology & Real-Time Intelligence
Mapping monolithic systems into interactive topological dependency graphs. Detecting circular imports, cyclomatic hotspots, and streaming real-time vector telemetry.
//Refracting Glass Prisms & Hallucination Sanitizers
When LLMs refactor code with tunnel vision, Prism parses the AST call-graph in <50ms to detect broken downstream callers, while Sober sanitizes leaked secrets and hallucinations.
//High-Speed Edge Routing & Parallel Filesystem Sweeps
Boundary-pruned crawlers reclaiming 50GB of dev bloat in 3 seconds, killing port hogs on :3000, and single-binary edge gateways routing inferences across 20+ LLM providers.
Experience the tools in action without leaving your browser. Click any quick-action or type directly below.
Architected for zero bloat, flow-state productivity, and automated AI pair programming.

Zero-Knowledge Local AI Distillation on Consumer Silicon
Cultivates massive foundation models into pocket-sized intelligences (0.5B-1.5B) directly on consumer laptops. Zero-OOM guard, prompt loss masking, and built-in blind holdout proof engine.

Developer Bloat & Zombie Dev Server Reaper
Parallel disk crawler that prunes traversal at artifact boundaries and hunts orphan dev servers holding ports :3000, :5173, :8000 hostage. Reclaims 50GB in under 3 seconds.



Autonomous Coding Agent & Pair Programmer
A next-generation multi-agent coding companion that thinks ahead, navigates deep codebases, learns developer style, and writes production code inside VS Code.

Autonomous Full-Stack App Builder
Give it a single prompt and watch it scaffold a production Next.js frontend, FastAPI backend, Dockerized orchestration, and GitHub Actions CI/CD pipelines instantly.

Universal Edge LLM Gateway
Single-binary, zero-config API gateway routing inferences across 20+ top-tier LLM providers with automatic fallback, load balancing, and token rate limiting at the edge.

Software Intelligence Analyzer
Reverse-engineers any codebase to generate interactive "DNA profiles", visualizing architecture, circular dependencies, cyclomatic hotspots, and developer patterns.

Modular Deep-Dive Research Agent
Intelligent CLI agent powered by LangChain that scours scholarly archives and the live web to generate synthesis reports with verifiable source citations.

Parallel DAG Workflow Engine
Executes highly complex data pipelines using strictly-typed Domain-Driven Design in Python. Includes micro-benchmarks, dependency graph validation, and task queuing.
A synchronized tripartite architecture combining autonomous multi-agent creation, sub-50ms AST contract radar, and zero-bloat runtime reapers.
Autonomous coding loops, contextual agents, and full-stack generative scaffolding that understand complex codebases.
Sub-50ms AST topological graph traversals to ensure AI agents never break downstream callers or leak credentials.
Parallel filesystem reapers and socket hunters that keep dev machines clean, freeing gigabytes of bloat and killing dead ports.
Autonomous agents construct complex software topologies from single prompts with verified domain models.
Languages, agent loops, and execution runtimes engineered with obsessive craft.
Colombo, Sri Lanka • Remote Global Architecture • Open-Source Creator (shenald-dev)
I am Shenal Diluksha (known across open-source as shenald-dev), a Senior Software Architect, systems designer, and creative vibe coder building zero-latency developer infrastructure, compiler AST invariant radars, and autonomous multi-agent AI systems.
My engineering philosophy is rooted in Zero Bloat and Maximum Flow: turning coffee and flow state into strictly-typed, sub-50ms execution engines. When AI writes code at 1,000 WPM, traditional tools introduce crippling drag, hallucinated bugs, and disk debris.
Rather than building trivial AI wrappers, I engineer foundational systems that empower developers: high-speed guardians like prism (sub-50ms topological blast radius radar), purge (parallel filesystem reaper & zombie server hunter), and sober (vibe coding hallucination sanitizer), alongside autonomous multi-agent platforms and edge gateways.
Self-directing agents that think ahead, write full-stack code, and synthesize deep academic research.
Topological graph modeling, circular dependency analysis, and quantitative market intelligence.
Sub-50ms topological call-graph guardians that detect AI refactoring breakage and slopsquatting.
Parallel filesystem sweepers reclaiming 50GB and single-binary sub-10ms edge LLM gateways.
Frequently referenced inquiries on engineering philosophy, open-source developer tooling, and technical invariants.
Shenal Diluksha is a Senior Software Architect, AI Systems Engineer, and Creative Vibe Coder. He specializes in designing deterministic software systems, microsecond developer tooling, and autonomous multi-agent frameworks that eliminate enterprise bloat and establish compile-time AST invariants.
Whether you need a principal software architect to design an autonomous multi-agent system, optimize high-throughput distributed pipelines, or build category-defining developer tools — let's talk.