# Pool Deza - Full Projects Technical Specifications (llms-full.txt)

This file contains the deep technical specifications, architectures, challenges solved, and business/technical impacts of the 12 primary projects in Pool Deza's portfolio.

---

## 1. ACUÑAPEDIA
- **Overview:** Intelligent Retrieval-Augmented Generation (RAG) system with auto-sync, hybrid search, and intent analysis for the political sector.
- **Architecture:**
  - Frontend: Next.js 15 (React 19, TypeScript 5.8) with Server-Sent Events (SSE) streaming support.
  - Backend: Python (FastAPI / LangChain) for embedding generation and query processing.
  - Vector Store: FAISS Vector DB hosting over 22,500 dense embeddings.
  - Relational Database: MySQL with real-time automatic synchronization pipeline.
  - Model: Grok AI 4.1.
- **Problem Solved:** Traditional search engines struggled to query across unstructured governmental plan documents, works databases, and live press releases simultaneously, resulting in poor accuracy and slow responses (8s+ latency).
- **Solution & Key Implementation:**
  - Developed a dense + sparse hybrid search integrating FAISS embeddings with BM25 keyword matching and a dynamic reranker.
  - Built an Intelligent Query Analyzer that identifies user intent to filter the search range and automatically adjust top_k retrieval weights.
  - Implemented real-time token streaming to lower perceived UI latency from 8s to just 1.2s.
- **Results / Impacts:**
  - 22,564 documents indexed.
  - Context precision accuracy improved from 77% to 89%.
  - Stable 1.2s streaming latency on standard AWS EC2 hosting without GPU requirement.

---

## 2. LEARNINGIFELSE
- **Overview:** Immersive Virtual Reality educational experience designed to teach programming logic (conditional branches) dynamically on Meta Quest 3.
- **Architecture:**
  - Core: Unity Game Engine & C# scripting.
  - Low-level: C++ / CMake native integration.
  - Modeling: Blender for low-polygon custom 3D asset optimization.
  - SDK: Meta Quest SDK with Hand Tracking integration.
  - Shaders: Custom HLSL / ShaderLab shader scripts.
- **Problem Solved:** Coding students often struggle with abstract computer science concepts (like conditional branches) when limited to text editors. Meanwhile, Quest 3's mobile chipset throttled standard 3D rendering, causing low framerates and motion sickness.
- **Solution & Key Implementation:**
  - Designed interactive coding challenges where students build logic maps with 3D physical nodes and interact using high-precision Hand Tracking.
  - Implemented an efficient Object Pooling system to reuse game actors, entirely eliminating CPU overhead and garbage collector spikes during dynamic respawns.
  - Crafted highly optimized HLSL shaders and baked lightmaps to maintain visual quality under mobile processing constraints.
- **Results / Impacts:**
  - Maintained a rock-solid, constant 60+ FPS on standalone Meta Quest 3 hardware.
  - 0% garbage collector spikes during run-time logic generation.
  - Dramatically improved logic comprehension metrics among testing engineering students.

---

## 3. CESAR ACUÑA 2026
- **Overview:** Immersive presidential campaign web platform with client-side AI, dynamic works geolocator, and high-performance scroll animations.
- **Architecture:**
  - Core: Next.js 15.5 & React 19 (TypeScript 5.8).
  - Styling & Animations: Tailwind CSS 4, GSAP (GreenSock ScrollTrigger), Framer Motion.
  - AI Engine: Local JSON-parsed Vector-less LLM ChatView client.
  - SEO & Analytics: Semrush, Google Analytics 4, native Progressive Web App (PWA).
- **Problem Solved:** Serving large-scale campaign portals to hundreds of thousands of users incurs huge backend compute fees, especially when integrating an interactive AI chat. Additionally, complex visual layouts usually lag on mid-range mobile devices.
- **Solution & Key Implementation:**
  - Built a local vector-less AI chat utilizing structured JSON representations of the candidate's government plans, run completely on the client-side to save on server costs.
  - Configured lazy loading, dynamic code splitting, and browser hardware acceleration via CSS `will-change` on GSAP triggers.
  - Built an interactive maps catalog linking 320+ candidate works with regional filtering and PWA install capabilities.
- **Results / Impacts:**
  - Scaled easily to support 100K+ monthly visitors with zero server-side AI computation cost.
  - Achieved perfect Core Web Vitals (Largest Contentful Paint < 1.2s).
  - 45% visitor-to-lead registration conversion rate.

---

## 4. FAVORAPP CRM
- **Overview:** Administrative control center and moderation dashboard with Turbopack for a high-traffic services marketplace.
- **Architecture:**
  - Core: Next.js 15, TypeScript 5.3, Radix UI.
  - Build Pipeline: Next.js Turbopack, Docker multi-stage.
  - State & Data: React Custom Hooks, optimized useMemo, REST API with JWT Interceptors.
- **Problem Solved:** Moderating and rendering 50K+ active/reported users and listings on standard client interfaces causes massive performance degradation and slow loading times.
- **Solution & Key Implementation:**
  - Implemented virtualization and O(1) in-memory lookups by optimizing filters with react-hook-form and react.memo.
  - Integrated Turbopack, accelerating local development compilation speeds 5x over Webpack.
  - Created a real-time user balance computation pipeline with Redis caching and event-driven invalidations.
- **Results / Impacts:**
  - Over 1,245 monthly reports moderated with 95%+ account validation accuracy.
  - Development HMR (Hot Module Replacement) compile times reduced under 2s.
  - 70% reduction in user moderation turnaround times.

---

## 5. FAVORAPP MOBILE
- **Overview:** Collaborative economy mobile application powered by a Go microservices backend.
- **Architecture:**
  - Frontend: React Native (TypeScript 97.9%, Kotlin 0.9%, Objective-C).
  - Backend: Go / Golang (99.6% coverage) consisting of 4 independent microservices.
  - Services: Core API (CRUD & Cloudinary upload), Messaging API (WebSockets & FCM), Identity (OAuth/Auth), and Sync Endpoints.
  - Storage: PostgreSQL, Redis.
- **Problem Solved:** Real-time messaging and push notifications in a serverless collaborative economy environment demand low latency and robust multi-device syncing while avoiding high-traffic server crashes.
- **Solution & Key Implementation:**
  - Programmed a low-latency WebSockets hub in Go leveraging optimized goroutines, connection pools, and automatic cleanup.
  - Integrated Firebase Cloud Messaging (FCM) with device token synchronization and multi-device fallbacks.
  - Formulated dynamic image compression uploading to Cloudinary asynchronously, bypassing blockages in the primary business API.
- **Results / Impacts:**
  - 4 production microservices supporting 10K+ daily active users.
  - Solid 99.9% uptime on the Go WebSocket messaging service.
  - Over 50,000 push notifications sent with zero delivery failures.

---

## 6. CENTRO ACADÉMICO POLÍTICO (JUVENTUDES APP)
- **Overview:** Digital education web portal featuring an isolated multi-role workspace architecture (Admin, Teacher, Student) and NextAuth v5.
- **Architecture:**
  - Core: Next.js 15, TypeScript 99.7%, Prisma ORM, PostgreSQL.
  - Auth: NextAuth v5 (Google OAuth, JWT Session persistence, automatic Refresh tokens).
  - Framework features: React Server Actions, Tailwind CSS.
- **Problem Solved:** Managing access control and tracking progress across three distinct user roles (teachers, admins, and students) usually results in severe state coupling and high database query counts.
- **Solution & Key Implementation:**
  - Engineered a fully decoupled 3-workspace modular routing structure, utilizing custom NextAuth middleware roles.
  - Replaced classic REST endpoints with React Server Actions and Prisma, pulling data directly on the server to reduce client-side fetching.
- **Results / Impacts:**
  - 101 commits of incremental clean-code development over 6 months.
  - 50% reduction in API overhead and database queries compared to traditional REST.
  - 0 reported authorization/RBAC security vulnerabilities in production.

---

## 7. SICARIO APP
- **Overview:** SaaS software performance and PC optimization marketing page and license portal.
- **Architecture:**
  - Core: Next.js 15, React 19, TypeScript, Tailwind CSS, Framer Motion.
  - Database: Prisma ORM, PostgreSQL.
  - Security & Auth: Auth.js v5 (NextAuth), bcryptjs, Zod validations.
- **Problem Solved:** Software subscription license systems require high-entropy cryptographic generation and secure administration, while the frontend must remain hyper-responsive and SEO-optimized.
- **Solution & Key Implementation:**
  - Wrote a secure license generator creating high-entropy unique key strings with immediate real-time activation checks.
  - Coded strict form state validations using Zod schema controls on both API routes and client screens.
  - Created a dual-currency payment structure (USD/MXN) with atomic updates to ensure data integrity during pricing edits.
- **Results / Impacts:**
  - 100% automated license verification and instant activation flow.
  - Secured administrative dashboards with complete protection against brute-force attempts.
  - Highly fluid UI driving high-intent customer acquisitions.

---

## 8. SISTEMA JUVENTUDES
- **Overview:** Enterprise resource planning and hierarchical organizational platform utilizing low-overhead real-time Server-Sent Events (SSE).
- **Architecture:**
  - Core: Next.js 16, TypeScript 5, Tailwind CSS 4.
  - Database: Prisma 6 ORM, PostgreSQL.
  - Real-time: Server-Sent Events (SSE).
  - Auth: NextAuth with custom Role-Based Access Control (RBAC).
- **Problem Solved:** Multi-tiered organizations (Regions, Provinces, Districts) have recursive, deeply nested hierarchies that cause N+1 database performance drops and lag during live resource updates.
- **Solution & Key Implementation:**
  - Avoided the high memory footprint of WebSockets by implementing Server-Sent Events (SSE) in Next.js, allowing native server-to-client streaming of calendar and chat events.
  - Modeled recursive Postgres relationships in Prisma 6 with highly optimized query indexing.
  - Wrote an automated global audit log system tracking all database mutations alongside the user's IP and timestamp.
- **Results / Impacts:**
  - Supports 500+ corporate users concurrently in active production.
  - Eliminated N+1 query lag, achieving sub-100ms response times on nested structures.
  - Robust auditing covering 100% of data mutations.

---

## 9. UNIVERSE DOWNLOADER
- **Overview:** Professional multi-platform media downloader and transcoder with AI Whisper transcriptions.
- **Architecture:**
  - Core: Python 3.11, FastAPI, FFmpeg.
  - AI Engine: OpenAI Whisper.
  - Build Pipeline: PyInstaller compilation.
  - Frontend: Vanilla JavaScript, HTML5, H.264 video standards.
- **Problem Solved:** Downloading, merging, and transcoding high-resolution video tracks (4K) freezes primary application interfaces, leading to memory crashes and unhandled system failures when social media APIs change.
- **Solution & Key Implementation:**
  - Constructed an asynchronous non-blocking download thread hub utilizing Python hooks to stream real-time speed, progress, and ETAs.
  - Built an automated ffmpeg container that automatically converts tracks to universal H.264 MP4.
  - Created an AI transcription pipeline using OpenAI Whisper to convert downloaded clips to text files on the fly.
  - Compiled the entire Python runtime and binaries using PyInstaller for an offline "Click & Run" standalone distribution.
- **Results / Impacts:**
  - High availability downloads up to 2160p (4K) from YouTube, TikTok, Twitch, Kick, and Facebook.
  - 100% success rate on multi-platform conversions with automatic fallback mechanisms.
  - Integrated click-to-run desktop distribution.

---

## 10. DIGITALBREAK
- **Overview:** Premium lead acquisition landing page and CRM platform with Core Web Vitals A+.
- **Architecture:**
  - Core: Next.js 15, React 19, TypeScript 5, Tailwind CSS 3, Prisma.
  - Security: reCAPTCHA v3, Content Security Policy (CSP) headers, CORS.
  - Forms: react-hook-form, Zod, Vercel SpeedInsights.
- **Problem Solved:** High-performance marketing campaigns experience lead leakage if forms load slowly or get spammed by bots, demanding absolute visual optimization alongside secure API submission.
- **Solution & Key Implementation:**
  - Implemented global React context translation maps to handle English and Spanish routes without duplicating components.
  - Secured the API using invisible reCAPTCHA v3 validations, restricting cross-origin inputs with strong CSP rules.
  - Optimized visual assets, utilizing next/image lazy loading, code-splitting, and clean Framer Motion transitions.
- **Results / Impacts:**
  - Captures 500+ qualified leads per month with a 15%+ conversion rate.
  - Verified A+ Core Web Vitals (Largest Contentful Paint < 1.5s).
  - Perfect 100% performance rating on Vercel Analytics.

---

## 11. ALIANZA PARA EL PROGRESO
- **Overview:** Official Peruvian political party institutional web ecosystem featuring a recursive resource center and bilingual Spanish/Quechua support.
- **Architecture:**
  - Core: Next.js 15, React 19, TypeScript 5, Tailwind CSS 3, Prisma ORM, MySQL.
  - Auth: NextAuth Google OAuth with custom role callbacks.
  - Multi-language: next-intl routing (ES/QU) with persistent locale context.
- **Problem Solved:** Organizing thousands of political documents, slides, and educational files into search-friendly archives while making the site fully accessible to native Quechua-speaking audiences.
- **Solution & Key Implementation:**
  - Integrated next-intl to translate both the UI and database queries dynamically according to the route.
  - Designed a nested category schema in Prisma with a recursive category tree search algorithm to index hundreds of files.
  - Configured lazy loading and dynamic skeleton loaders on the congressional grid, accelerating first load speeds.
- **Results / Impacts:**
  - Handles 10K+ monthly active visits with a solid 99.9% uptime.
  - Ranked #1 on Google Peru search results for political queries.
  - High-engagement newsletter capturing over 2,000 active subscribers.

---

## 12. SOCIAL MANAGER
- **Overview:** Social media administrative CRM with real-time Graph API syncing, Socket.IO WebSockets, and AI sentiment analysis.
- **Architecture:**
  - Frontend: Next.js 16, React Server Components, TypeScript 99.3%, Tailwind.
  - Backend: NestJS 11 modular enterprise architecture (TypeScript 99.5%).
  - Real-time & Caching: Socket.IO Gateway, Upstash Redis, NestJS Task Scheduler.
  - API: Facebook Graph API, Webhooks, AI sentiment engines, Passport JWT.
- **Problem Solved:** Processing hundreds of live Facebook comments, detecting their emotional sentiment, and rendering them in real time demands high compute efficiency and secure API authorization.
- **Solution & Key Implementation:**
  - Structured a modular NestJS backend separating key domain blocks (Auth, Posts, Comments, Analytics, Realtime, AI).
  - Secured Facebook Webhook triggers by intercepting the raw payload and validating the HMAC-SHA256 signature.
  - Deployed Socket.IO with Redis adapters to allow scalable WebSockets, with a fallback to memory in local environments.
  - Scheduled automated batch syncs every 3 minutes to update metadata in the background.
- **Results / Impacts:**
  - Reduced manual social media analysis time by 70%.
  - Stable automated sync cycles running every 3 minutes.
  - Real-time AI sentiment analysis coverage on 100% of user comments.