Armando Peña Tamayo
Senior Full-Stack AI Engineer | SaaS Architecture | LLMs & RAG
About
Senior full-stack AI engineer with 8 years of experience designing and delivering production software end to end, from architecture and data modeling to implementation, integration, and deployment. I build multi-tenant SaaS and production-oriented AI systems: LLM integrations, RAG and vector databases, multi-stage agent orchestration, custom tools, and deterministic validation around probabilistic outputs. I work remotely with international engineering teams and treat reliability, security, and concurrency as core architectural concerns.
Experience
-
CBS Cleaning
Senior Full-Stack Developer
-Production-oriented AI automation platform for a document-heavy email back office. Built a hybrid deterministic/LLM pipeline: deterministic ETL and normalization, multi-format document extraction (including image-based text), Qdrant-backed RAG over indexed email documents and attachments, and multi-stage n8n agent orchestration with Structured Outputs and tool calling. Agents retrieved geographic and tax information through custom tools, selected OpenAI models by task complexity (GPT-4.5 and smaller models), and shared temporary per-email workflow state through Redis. Deterministic validators using the same APIs cross-checked low-confidence results. Added retry/recovery paths, concurrency controls, and duplicate/conflict prevention for SharePoint writes while processing hundreds of emails daily—thousands during initial rollout—cutting manual processing time by ~70%.
-
SuitUp Project
Senior Full-Stack Developer
-Mobile-first body-measurement and virtual try-on MVP for fashion e-commerce. Owned the full-stack build: Next.js architecture, camera-based capture flow, measurement processing, and AR try-on integration. Shipped in four months, reducing time-to-market by ~50% versus typical product timelines.
-
AmerAdjust
Senior Full-Stack Developer
-AI document-generation platform for the insurance industry, built from the MVP stage. Designed and implemented multi-tenancy from scratch: organization tenant IDs on business records, Supabase RLS enabled by default for new business-data tables through migrations, tenant-scoped hierarchical RBAC, and defense-in-depth authorization across database RLS and application middleware. Tenant administrators can access all users within their own organization; global administrators can operate across tenants; ordinary users and organizations remain isolated. Also built the Next.js/Supabase product and AI document pipeline, reducing document creation from hours to minutes (~80%). Implemented at MVP stage; multi-organization behavior was not yet validated in production.
-
Events Management SaaS
Senior Backend Developer
-Serverless microservices for an events-management SaaS on AWS Lambda. Built and maintained backend services and centralized logging/monitoring across functions, reducing incident debugging time by ~40% and improving reliability.
-
Explora Jalisco
Full-Stack Developer
-Admin and customer-facing Angular applications integrated with a CMS and backend services. Optimized SEO and frontend performance, improving page load speed by ~30%.
-
Wassermeloni
Frontend Developer
-Modernized a legacy Angular and Ruby on Rails platform and its delivery workflow. Introduced CI/CD and automated testing to stabilize releases and reduce deployment costs by ~50%.
-
Imagined Earth
Frontend / Mobile Developer
-Rebuilt a mobile app with Ionic and Angular, improving client performance and backend/API efficiency. Reduced server costs by ~30% and improved user retention through a faster, more reliable experience.
-
Freelance Projects
Full-Stack Developer
-Delivered end-to-end web applications for remote clients across LATAM, including event systems, remittance platforms, and HR tools. Owned requirements, architecture, implementation, authentication, and deployment from concept to production.
Education
-
University of Informatic Sciences
Skills
Languages
Projects
-
AI Job Hunter — Personal Agent System
Independent personal project. Autonomous agents research the web to discover and extract job opportunities, normalize listings, and evaluate eligibility with custom tools plus deterministic rules for geographic restrictions, payment-method compatibility, role type, and technical requirements. Multi-stage orchestration combines structured job processing and LLM reasoning with rule-based validation, then adapts the CV and generates a tailored cover letter for each position.
-
AI Security Testing Agents — Personal Project
Personal AI-assisted application security testing toolkit. Orchestrates agents to run active scans with Burp Suite Community Edition and OWASP ZAP, apply OWASP-guided checks and vulnerability checklists, and surface findings for review during development. Complements manual validation and professional penetration testing; it does not certify security or guarantee vulnerability detection.
-
adev-lru
TypeScript library implementing an O(1) LRU cache with TTL and persistence layers for LocalStorage and IndexedDB. Published on npm.
www.npmjs.com/package/adev-lru -
tkt
Go CLI that converts TODO comments into tickets, manages git branches, tracks time, and integrates with Neovim through a real-time file watcher.
github.com/Armando284/tkt -
Collaborative Whiteboard
Low-bandwidth real-time collaborative drawing app optimized for unstable networks and Apple Pencil, using WebSockets and RLE compression.
whiteboard-five-gamma.vercel.app -
Life Simulation
Neuroevolution simulation combining neural networks and genetic algorithms; virtual creatures evolve survival behaviors in a dynamic, canvas-rendered environment.
life-simulation-self.vercel.app -
Taller Web CU
Personal blog and services page for freelance web development, technical writing, and project case studies.
tallerweb-cu.vercel.app