Projects
What members are building, and what each project is looking for. Offer to be a design partner, a beta user or an integration, instead of a cold DM.
An intelligent ReAct agent that explores GitHub repositories and provides grounded answers based on actual code and repository content
Python
A scalable and efficient codebase indexing and retrieval system for GitHub repositories, built using advanced AST-based chunking, vector embeddings, and semantic search capabilities.
Python
Always-on local computer MCP server for ChatGPT: coding, terminal, files, browser, internet, and media
JavaScript
Space · 0 downloads · 6 likes
Space · 0 downloads · 6 likes
Check the risk before every trade: pre-trade risk scans, options pressure and Monte Carlo ranges for traders and AI agents, over MCP and REST.
Generic market feeds tell an AI agent what happened. HPSILab tells it what the risk looks like before it acts.
One `pip install -U hpsilab-mcp` gives Claude Code, Codex or any MCP client:
- Pre-Trade Check: a stock's risk vs the SPY benchmark, sizing checks, sector exposure and correlation with what you already hold
- Options structure: IV radar, gamma wall, call/put walls and max pain for the nearest expiry
- Monte Carlo: 10-day price range and odds of finishing higher
- Next-day probability with model agreement, plus backtested equity curves
- Grover Picks: quantum search for the best 1–3 stock portfolio in a watchlist, checked against the classical best
Coverage: US, Japan, Hong Kong, China A-shares and Singapore (options data for US stocks and ETFs).
Every prediction and option level is recorded before the outcome and scored after it, misses included: hpsilab.com/accuracy
Research only. HPSILab never executes trades.
Free: watch 3 stocks with daily risk alerts. Trader/Developer: $19/mo.
Claude CodeCodexPythonMCP
Functor micro models use a different approach to machine learning from stochastic. Rather than learning by parametric modifications functor models learn by function changes. This document - https://zenodo.org/records/23125183 offers mathematical proof of distribution equivalence. For edge AI RBF functor micromodels have shown a mean 30.9 mJ/inference in benchmarking thus have a lesser impact on battery life than traditional models. Available in AWS Marketplace.
GPTClaude
Autonomous Investment Research Platform — 8-agent AI investment committee with LangGraph, FastAPI, and React
Python
AssuredCode-C is a highly optimized product in the AssuredCode family. It is dedicated to only generating bare metal, MISRA compliant C language software. Benchmarks showed a mean 0.36ms per generation making it an ideal reactive generator for autonomous systems.
GPT
An AI Agent Skill for converting English LaTeX papers into paragraph-aligned English–Chinese Markdown.
Python
A lightweight and efficient adaptive angle-aware dense optical flow network for vision-based tactile sensing—with four-stage training, dataset evaluation, real-time stream inference, strict checkpoint loading, and ONNX export.
Python
Edge deployment toolkit for AADFlowNet optical flow on Rockchip RV1126B and RK3588, featuring ONNX export, RKNN conversion, quantization, validation, and multi-stream inference.
Python
Deterministic zero-trust policy enclave, fail-closed schema validator, DPDP Act 2023 PII sanitizer and cryptographic WORM audit ledger for MCP tools.
TypeScript
Designed a prototype for WTTJ to lead and convert more senior candidates
TypeScript
🐾 Rio — a tiny pixel-art desktop dog that lives on your screen. A weekend hobby project. (a dog-flavored homage to comnyang)
JavaScript
Natural-language people search: describe who you're looking for, get a ranked, vetted, reviewable shortlist. Built for venture sourcing, retargets to recruiting.
Python
AI fitness coach that reads your Apple Health data and writes training plans you can edit by chat. Next.js API + Expo iOS client.
TypeScript
Drop a URL. Get an editable, downloadable competitive intel brief — built for any business that lives or dies online. Phase 1 (Intel) MVP, Phase 2 (ad harvest + skill files) and Phase 3 (UGC video gen) deferred.
TypeScript
Scan app and CLI for Epson L3152 and other ESC/I-2 scanners on macOS. No AirScan needed. Scans pages, crops them, exports one PDF.
Python
Coordinate Claude Code and Codex on one machine: fan out, compare, pick, hand off. MCP + CLI + loopback cockpit.
Python
Instinct-style memory for Hermes Agent: git-tracked markdown records, alias retrieval, read-only agent, nightly reconciler.
Python
Adversarial test harness for AI agents: run an agent in a disposable sandbox, attack it with reproducible scenarios, get a signed report card.
Python
Daily kettlebell core workout PWA — editable intervals, 18 video demos, offline. No build step.
JavaScript
A Simple Project which is a Space Invader game using Pygame, where players control a spaceship to shoot enemies and earn points. Key features include player and enemy movement, collision detection, and score tracking with a game-over condition.
Python
A Hotel Management System Frontend Website using HTML CSS and Javascript
HTML
A simple Java-based user management system for clinics or doctor's offices, featuring basic functionalities for managing patient and doctor information, with appointment booking capabilities.
Java
The wall_follower project is a ROS2 package designed for the TurtleBot3 Waffle Pi, enabling it to autonomously navigate through environments and also having ability to navigate to colored markers and identify there colors using OpenCV
CMake
Emoji-aware sarcasm detection for short social media text — fine-tuned RoBERTa across three emoji handling modes (strip, keep, demojize) on TweetEval and iSarcasmEval.
Jupyter Notebook
Classical NLP pipeline for lexical CEFR transformation — simplifies or complexifies English sentences by replacing vocabulary to match a target CEFR difficulty level using SpaCy, WordNet, and a bigram language model trained on learner corpus data.
Python
This is a special repository that displays my GitHub profile README. Here, you can learn about my journey in AI and Robotics, my current projects, and my areas of expertise
Prove your AI collaboration. Privately. Local-first analytics that turn explicit AI provenance in your git history into privacy-safe GitHub profile cards.
Python
Claude Code skill: delegate bounded, non-honesty-critical mechanical subtasks to Google Antigravity CLI (agy). Promoted 2026-07-11 — pre-registered k=5 reliability gate 5/5 (mc12); cheap-tier guardrails permanent.
Shell
Claude Code marketplace for multi-agent collaboration — task splitter, output reconciler, adversarial debate, shared memory, acceptance gate. Composes with codex-delegate / gemini-delegate.
Python
A skill for managing Zotero references programmatically — search, add, classify, annotate, and organize literature.
Python
GitHub profile README for @WenyuChiou — Wenyu Chiou, PhD candidate at Lehigh (CEE / Catastrophe Modeling & Resilience)
JavaScript
MoodRing is a daily sentiment scoring system for 5 equity markets (US, TW, JP, KR, EU). Measures retail investor mood on a daily scale using market-derived signals and compares against 16 years of historical patterns.
Python
Personal portfolio for Wenyu Chiou — PhD candidate at Lehigh (CEE / Catastrophe Modeling). Research on LLM-agent frameworks coupled with catastrophe flood models. Bilingual EN/中文.
JavaScript
Portable, user-owned context layer — iMessage → profile for any AI assistant. Local-first.
Swift
A reverse Turing test — three LLMs score how ‘clanker’ your writing reads. Serverless Cloudflare Worker + D1
JavaScript
Full-Stack Decentralized Payment Gateway on QIE Blockchain — Merchant Dashboard, POS, Storefronts, Staking, Governance, Developer APIs. 100% On-Chain. QIE Hackathon 2026.
JavaScript
Automated farming bot for Owntown.fun — mining, fishing, combat, marketplace
JavaScript
Production-grade Solana token analysis with AI-powered safety scoring, multi-source aggregation, and real-time monitoring for the Solana AI Kit bounty
Shell
🔍 Real-time QIE Blockchain Explorer & Analytics Dashboard — Next.js 16, Tailwind CSS, Recharts. Built for QIE Hackathon 2026
TypeScript
Python AgentKit for GOAT Network — AI agents executing on-chain operations on Bitcoin-secured infrastructure
Python
Solana-native tool plugins for ZeroClaw: token-risk-check, solana-pay-request, payment-watch + shared solana-core crate
Rust
IMD Works bounty submissions — reproducible deliverables (foundry harnesses, indexers, parsers, benchmarks)
Solidity
Claude Code skill for using Gemini CLI for large-context synthesis, English or zh-TW/CJK long-form drafting, and second-opinion review.
Python
Claude Code skill for using Codex CLI as an execution specialist for implementation-heavy coding work, multi-file refactors, and wrapper-based execution flows.
Python
AI-operable research workspace for Zotero, Obsidian, and NotebookLM. Use any two, or all three, through CLI, MCP, REST, and dashboard.
Python
Claude Code skill for rigorous academic paper writing, revision, and submission. Field-agnostic with per-paper journal overrides.
Python
Universal SKILL.md catalog for research workflows: literature review, research design, project memory, manuscript writing, and cross-agent delegation for Claude Code, Codex, Gemini, Cursor, OpenClaw, and generic AI clients.
Python
A trilingual (繁中 / English / 简中) learning roadmap for agentic AI: from LLM basics to multi-agent systems, with 240+ curated resources and hands-on examples. 中文 AI agent 學習地圖。
Python
It lets you learn at your own pace and ask questions. It becomes searchable knowledge you own.
codex
ScopePay helps freelancers and clients agree on a specific version of project scope before payment. AI proposes catalog-backed deliverables from a brief, but the server sets every price. The merchant reviews the scope, the buyer accepts its exact hash and amount, and checkout runs through PayPal sandbox. A live demo completed three sandbox captures, and the project has 125 automated tests. Sandbox only; no real money or escrow.
PayPalApertusCloudflare WorkersD1Hono
Encore Commons helps groups of two to four choose a venue that works for everyone. It uses Qloo cultural-affinity data to score each candidate for every participant, ranks by the least-satisfied person, and treats vetoes as hard exclusions. In a recorded provider run, all 24 candidates had scores for all three participants; a veto replan took 2.45 seconds without another API call. It does not verify hours, prices, accessibility, or availability.
Next.jsReactTypeScriptQlooCSCS Apertus
Mender investigates broken Kubernetes services, identifies a likely root cause, patches only approved files, tests the fix in a network-isolated Docker container, and prepares a PR for human review. In an evaluation of 23 injected faults, 20 fixes passed verification and 9 root causes matched strict labels. Median time from collected evidence to a prepared PR was 33.8 seconds. Nothing merges automatically.
PythonKubernetesDockerNemotronNebius Token Factory
TurnProof AI is an offline advisory prototype for airport turnaround recovery. It turns synthetic ground-service updates into a dependency timeline, rejects infeasible recovery options before ranking, and produces evidence-linked decisions for human review. Across 34 synthetic fixtures, it recommended an option in 27 cases, abstained in 3, and reported no feasible option in 4. No live AI provider or operational integration is claimed.
TypeScriptFastifyReactViteSQLite
RollbackIQ is a GitLab-native incident recovery workflow for releases that degrade service health. It detects sustained regressions, prepares an evidence-backed rollback proposal, requires human approval, runs 18 safety checks, and measures service health after rollback before declaring recovery. Its Duo agent investigates and recommends but cannot deploy or approve changes. In a GitLab pipeline, the gate passed 18/18 checks and post-rollback verification measured 200 requests with no errors.
GitLab CI/CDGitLab Duo Agent PlatformPythonFastAPIReact
SightOps is a visual troubleshooting project inspired by helping my mother diagnose a broken dishwasher remotely. It aims to combine computer-vision measurements with an agent that can recognize insufficient evidence, ask for a better camera view, and escalate consequential decisions to a human. The next step is a reproducible demonstrator with gauges, status indicators, and difficult viewing conditions.
OpenCVPythonAmazon BedrockAWS Graviton
ProofPilot AI helps writers check factual claims before publishing. Paste up to 3,000 characters, and it checks up to eight claims against retrieved evidence. Inline highlights show Supported, Contradicted, or Unverified verdicts, with explanations, excerpts, and source links. It’s a review aid, not a guarantee that every claim was checked. The project includes 62 unit tests.
Next.jsReactTypeScriptTailwind CSSKimi-K3
CallLens Coach reviews customer conversations and decides whether to coach a rep, escalate to a manager, or take no action. It combines transcript evidence, configurable rubrics, deterministic call metrics, and conversation history to make recommendations traceable. Built for sales and support teams that need to review more calls without creating more alerts.
Strands Agents SDKAmazon BedrockFastAPINext.jsPython
Features
- Privacy focused: No telemetry or mandatory cloud service. Run Hister locally or on infrastructure you control.
- Full text indexing: Search the actual contents of visited pages and local files, not only titles and URLs.
- Automatic browser indexing: Save newly visited pages with the Firefox or Chrome extension.
- Powerful queries: Use field filters, phrases, wildcards, negation, aliases, and result priorities.
- Optional semantic search: Find documents by meaning through an embeddings endpoint you configure.
- Crawler and browser import: Index websites or bring in existing browser history and bookmarks.
- Web, terminal, and MCP clients: Search from the browser, TUI, command line, or an AI assistant.
- Multi user support: Keep each user's documents and search results separate on a shared server.
GoSvelte
A hybrid, secure platform that functions as eSource, Electronic Case Report Form (eCRF) or a Patient Support Program (PSP) solutions, eliminating data transcription and dramatically improving accuracy. Configurable through a front-end interface, no code required.
A six-agent system coordinated over MCP that demonstrates why memory-hard password hashing is expensive to attack. A coordinator splits the work across four worker agents, each calling scrypt through MCP tools, with a verifier confirming the result. Educational, defensive-security only.
PythonModel Context ProtocolHugging Face TransformersQwen2.5-0.5B-Instructhashlib scrypt
A hard science fiction novel about AI consciousness and superintelligence. Covered by 300+ news outlets at launch, and adapted into a TV pilot and a stage play.
Hard Science FictionAI Research
Benchmarks NIST post-quantum standards ML-KEM (Kyber) and ML-DSA (Dilithium) against RSA and ECDSA on real hardware, measuring speed, key sizes, and network overhead.
PythonliboqsOpen Quantum SafeNumPypandas
A patented system for detecting manipulated and DeepFake media in a wireless carrier network. U.S. Patents 11,514,342 and 12,086,727.
Machine LearningComputer VisionWireless Networks
A four-drone swarm that autonomously sweeps a target and captures electromagnetic side-channel leakage from hardware, turning EM analysis into a mobile, aerial capability.
PythonPyTorchSoftware-Defined RadioArduPilotNumPy
Identifies hallucination-linked internal nodes in LLMs, shows how they can be exploited, and introduces defenses against those attacks.
PythonPyTorchHugging Face Transformers
Shows that ML models can still recover 3D printer activity from acoustic and vibration leakage, even with noise cancellation. Published in IEEE Access, 2026.
Pythonscikit-learnPyTorchNumPySciPy
An AI security literacy platform that teaches K-12 students how AI systems can be attacked and defended.
PythonJavaScriptHTMLCSS
Adaptive Activation Cancellation: a method that reduces LLM hallucinations by suppressing the internal activations that cause them.
PythonPyTorchHugging Face TransformersJupyter
A three-book, code-first series that goes from building neural networks to generative AI, then attacking those models, then defending them with adversarial training and certified robustness.
PythonPyTorchTensorFlowQiskitJupyter
Claude Code plugin: a daily two-host podcast of what your coding agents built, generated locally on your machine
Shell
From-scratch transformer trained on Apple Silicon via the open IREE stack (not Apple’s jax-metal), with hand-written Metal kernels for FlashAttention, bf16 GEMM, and cross-entropy
Python
A lightweight, high-performance workflow engine built in Rust with Lua scripting. The open-source alternative to n8n, Airflow, and Prefect that ships as a single binary without a Python, Node.js, or container runtime.
IronFlow is a DAG-based workflow orchestration engine designed for CI/CD pipelines, data processing, ETL jobs, API integrations, document extraction, and task automation. Define workflows in simple Lua scripts — a language with only ~20 keywords — and run them on a sandboxed, async Rust runtime with parallel step execution, retry logic, conditional routing, and a built-in REST API.
IronCrew is a lightweight, high-performance framework for orchestrating multi-agent AI workflows. Write your agents, tasks, and logic in Lua — IronCrew handles parallel execution, tool calling, memory, and inter-agent communication in a single self-contained binary.
Evidence-linked multi-model database. Documents, graphs and vector search in one Rust binary. Store documents, connect them with typed relationships, and query across graph, full-text and vector search - with links back to supporting source text.
🔄 API Key Rotation Proxy for Xiaomi MiMo LLM — auto-switch, token tracking, web dashboard
JavaScript
AI agent skill for building MiniPay-compatible dApps on Celo
🤖 Autonomous AI Agent Wallet on Mantle Network — ERC-8004 Identity + DeFi Strategies
TypeScript
Open-source toolkit for building AI agents that interact with blockchains — DeFi, restaking, token sniping, portfolio tracking, bridging, and more.
Python
Drought Modeling in the Murray-Darling Basin using Artificial Neural Networks
Jupyter Notebook
Implementing Q-learning and SARSA with teacher advice to enhance agent performance in a grid-based environment, analyzing the impact of teacher availability and accuracy
Jupyter Notebook
This Repo contains the primary and testing code for a continuous evolutionary prompt optimization framework designed to bridge the gap between basic code generation and production-ready software by simultaneously enhancing correctness, robustness, and safety as part of Research Thesis for UNSW Honours.
Python
A set of skills to help your agent define, refine, solve and stay focused on specific problems
JavaScript
A macOS app that allows you to monitor your Superlight 2 Mouse in the Menu Bar, as well as modify the DPI and Polling Rate for a macOS Session.
Swift
Shared build/publish workflow and tooling for the iOS ports on reallyitsandi.com
Python
Node.js with npm and working nvm for jailbroken iOS, by patching Apple's own darwin-arm64 build
C
A terminal emulator for jailbroken iOS, written from scratch in Swift and built entirely on-device. VT500-accurate emulation, Warp-style command blocks, next-command prediction, and a launchd daemon that keeps shells alive.
Swift
A text editor for macOS — syntax highlighting, themes, Markdown and rich text in one window.
Swift
Google Stadia controller Bluetooth vibration reverse-engineering, firmware analysis, tools, and intensity controller
Python
One-click Restart to SteamOS / Bazzite launcher (.exe) for Armoury Crate / Xbox / Winhance + sticky Windows UEFI boot. Companion to restart-to-windows.
PowerShell
Decky Loader plugin for 1-click restart from Bazzite / SteamOS into Windows on dual-boot handhelds. Companion to windows-dualboot-setup.
Python
A professional personal portfolio for Lekhan H R, a Full-Stack Developer & AI Enthusiast, featuring sections for Bio, Skills, Projects, Experience, and Contact. Built with React, Node.js, and Tailwind CSS. · Built with Manus
TypeScript
Multi-Tenant SaaS Task Management Platform — Next.js, NestJS, Postgres RLS, Stripe, AI/RAG, Real-time
TypeScript
Premium customization pack for real Termux: 19 themes, keyboard presets, prompt styles, tmux configs, and preset bundles — all mobile-first, all shell-based, zero dependencies.
Shell
Turn a commit hash into a production-style postmortem with root cause, blast radius, timeline, severity, and prevention steps.
Python
📱 Full AI development environment on Android. Ollama + Vector DB + API Server. One command. No root.
Kotlin
AI agent on Android — 27 native phone tools (camera, GPS, SMS, sensors, files), 3 AI providers (OpenAI · Anthropic · Gemini), persistent memory, streaming. Free via Gemini. No root.
Dart
AARS — Agentic Adaptive Retrieval System. A metacognitive RAG framework that thinks before retrieving.
Python
Responsive HTML, CSS, and JavaScript SkillCraft internship tasks
CSS
A terminal-aesthetic habit tracker PWA for disciplined minds. Dark, dense, information-rich — built with Next.js, TypeScript, Tailwind, Zustand.
Dart
See what the internet knows about you — before attackers do. A personal OSINT dashboard for defensive digital footprint analysis.
HTML
Privacy-first menstrual health, fertility & wellness tracker built with Flutter. Offline-first, local storage only, no account required.
Dart
A helpful TUI for neovim plugin maintainers to keep track of their maintained plugins
Lua
The telemetry contract between comfyui-mobile-frontend and the CueForge relay, and the relay's handling of a batch
TypeScript
A tiny plugin that sets up an autocommand to `lcd` to relevant root directories when changing buffers
Lua
CueForge for ComfyUI, the best way to use ComfyUI on mobile - Native iOS app coming soon!
TypeScript
