I'm a Peruvian Fullstack & AI Developer with a Bachelor's Degree in Computer Science from Peru. I specialize in building modern, scalable web applications and integrating Intelligent AI Agents, RAG architectures, and workflow automation.

My fullstack tech stack centers on Node.js, Express, JavaScript, TypeScript, React, Redux Toolkit, FastAPI, Python, HTML5, CSS3, and Tailwind CSS, backed by strong data handling with Zod and Sequelize/PostgreSQL. Beyond traditional web software, I design smart ecosystems leveraging LLMs, decision graphs (LangGraph/LangChain), and process automation (n8n).

I am a continuous learner who continuously sharpens technical skills through platforms like freeCodeCamp, Frontend Mentor, YouTube, adventJS, and HackerRank. I graduated from the ONE (Oracle Next Education) program as a frontend web programmer with React. Since then, I have actively delivered team projects in professional work simulations like No Country and Foo Talent, as well as working on custom freelance solutions.

• I am interested in Web Design & Development, AI Agents & Automation, and QA Testing.

• 🌱 I’m currently deepening my knowledge in:
  • Java & Spring Boot
  • Python (FastAPI & AI Orchestration)
  • Angular
  • AI Frameworks & Agents (LangChain / LangGraph / n8n)
  • Figma (UX/UI Research & Prototyping)
• 👯 I’m looking forward to collaborating on open source, Fullstack, and AI-driven projects.
• ✔ Ask me about anything, I am happy to help, only if the ball is in my court! 😉
  
• Outside tech, 📖 I love to read novels, watch movies, 🎵 listen to music, and 🌴 explore nature outdoors.

• 📧 Gmail Primary: limayolivamary@gmail.com
• 📧 Outlook Alternative: marycarolimay@outlook.com
• 💼 LinkedIn: linkedin.com/in/carolina-limay-oliva

Solid knowledge in User Experience (UX) Design principles and methodologies

Leveraging AI architectures, LLM frameworks, and tooling for intelligent automation best practices

Project developed for the Kiro 2026 Hackathon (organised by Código Facilito & AWS)

Description: An AI-driven technical coaching platform that works directly with your real code. It analyzes GitHub repositories, generates tailored improvement tickets, and conducts interactive technical interviews (via chat or voice) to evaluate and provide feedback on your software decisions.

Key Features:

• 🤖 4 Specialized AI Agents Pipeline: Multi-agent orchestration powered by Gemini 2.5 Flash featuring Code Reviewer, Ticket Generator, Tech Lead, and Evaluator.
• 🎙️ Real-Time Voice Interviews: First platform to conduct voice-based code evaluations using native Web Speech API.
• 📊 5-Dimensional Evaluation: Comprehensive scoring across technical understanding, justification, alternatives, limitations, and communication skills.
• 🎮 Gamification & Dashboard: Features 11 levels, XP points, streaks, 8 unlockable badges, and a global leaderboard.
• ☁️ Cloud Infrastructure: Fully deployed on AWS (S3, CloudFront, Elastic Beanstalk) and Supabase (PostgreSQL), managed via Terraform.

Technologies Used: React 19, TypeScript, Redux Toolkit, Tailwind CSS, FastAPI, Python, Gemini 2.5 Flash API, Supabase (PostgreSQL), AWS (S3, CloudFront, Elastic Beanstalk, Secrets Manager, CloudWatch), Terraform.

Description: A full-stack intelligent assistant for the online school "TechAcademy", built with a RAG (Retrieval-Augmented Generation) agent that answers questions about courses, institutional policies, and academic processes. The system combines vector search over official documentation (PDFs) with structured queries over tabular data (CSV), using a LangGraph decision graph that routes each question to the most appropriate information source.

Key Features:

• 🧠 Multi-Node Decision Graph: LangGraph workflow with 6 specialized nodes (Triaging, Pandas Agent, RAG, Clarifier, Guardrail, Formatter) that intelligently routes each query to the optimal data source.
• 🔍 Hybrid Knowledge Base: Combines FAISS vector search (MMR, top-k=5) over PDF documentation with a Pandas Agent for structured CSV queries (prices, schedules, professors, vacancies).
• 💬 Conversational Memory: Injects the last 6 messages as context so the agent resolves ambiguous references like "how much does it cost?" when a course was previously mentioned.
• 🛡️ Guardrail & Clarifier Nodes: Off-topic questions are politely redirected; ambiguous queries prompt the user for clarification instead of guessing.
• ⚡ Optimistic UI & Background Requests: Messages appear instantly; if the user switches threads while waiting, the response is saved to the correct thread without affecting the current view.
• 🌗 Full-Featured Chat UI: Dark mode, responsive design, animated typing indicator, auto-generated thread titles, JWT authentication with Argon2 hashing.

Technologies Used: Python 3.11, FastAPI, LangChain/LangGraph, Google Gemini 2.5 Flash, FAISS, Pandas, SQLAlchemy, Alembic, PostgreSQL (Supabase), JWT + Argon2, React 19, TypeScript, Redux Toolkit, Tailwind CSS 4, Vite, Axios, Zod.

Deploy: Render (Backend) + Vercel (Frontend)

Description: An intelligent AI Agent built to automate Human Resources workflows and instantly manage corporate inquiries like vacation balances, time banks, and work modalities. Based on the core framework from the Alura Latam Immersion, this system was enhanced to meet strict enterprise standards for data privacy and backend logic.

Key Features:

• 🔒 Biometric Identity Filter: Uses unique telegram_id validation via Supabase to automatically allow or block data rendering before the LLM processes private queries.
• 🛡️ Anti-Spoofing & Auto-Link: Prompts unknown users for their DNI, parses digits syntactically with AI functions, and securely pairs accounts only if the document is free, locking out identity hijackers.
• ⏳ Dynamic Time Anchoring: Injects automated, real-time timezone context (America/Lima) into the prompt payload so the agent computes employee tenures and dates flawlessly.
• ⚙️ Hybrid Multi-Tool Architecture: Orchestrates LangChain tools so the agent dynamically selects between an In-Memory Vector Store (RAG) for general company rules and relational tools for PostgreSQL updates.

Technologies Used: n8n (Advanced AI Agents), Cohere LLM, LangChain Framework, Supabase (PostgreSQL), Telegram API.

Project	Description	Technologies	My Role	Links
Valhala	Liquor store management system: Backend for inventory control and sales tracking with admin/worker roles. Future e-commerce platform (frontend in development).	Node.js, Express, Sequelize, Zod, JWT Authentication	Backend Developer
FastLab	Clinical Laboratory Management System: A web application for managing medical laboratories, including test results, patient information, and medical staff administration.	React, Redux toolkit, JavaScript, C# .NET	Frontend Developer
ServiHogar	Home services hiring platform: MVP of a marketplace for finding and hiring domestic service providers, where anyone can also register to offer their services. Mobile-only MVP: Application designed specifically for smartphone use cases	React, Redux Toolkit, TypeScript, Node.js, Express, MongoDB	Frontend Developer
Trucking Management	Logistics & fleet management system: MVP of an app for managing the vehicle fleet of an entire company	React, Redux toolkit, JavaScript, Java Spring Boot, MySQL	Frontend Developer
GreenIX Platform	Online store for eco-friendly and sustainable products: MVP of an e-commerce about sustainable products	React, Node.js, JWT Auth	Frontend Developer

All projects developed in collaborative team environments using Agile methodologies

Description: A mobile application designed to manage and track your pet's complete medical history. The app allows pet owners to record veterinary visits, consultation reasons, treatments, and medications. The medical records can be easily shared with veterinarians, pet sitters, or family members.

Key Features:

• 📋 Complete medical record tracking
• 🏥 Veterinary visit logging with reasons
• 💊 Medication and treatment history
• 📤 Easy sharing of medical records
• 📱 Mobile-first design for pet owners on the go

🚧 Portfolio website coming soon! Meanwhile, check my GitHub projects below.