AI Automation & Backend Development

Donato Hernández

AI Automation · Backend Systems · Systems Integration

I build reliable systems that take the repetitive work off your plate.

From workflow automation to live AI conversations, I build the backend and integrations that make your tools work together. My stack includes Node.js, TypeScript, Python, n8n, and Google Cloud.

View ProjectsLet's Connect
CV
Donato Hernández, AI automation and backend developer
Donato Hernández
01 · About

Better workflows. Reliable systems.

I build software around the way people actually work. For a family-owned business, that meant replacing handwritten credit records with a mobile workflow. For PitchPilot, it meant making live AI conversations feel responsive and useful. With Zaaby, it meant helping people put saved content to use. I start by understanding the process, then design the data, integrations, and backend that make the solution reliable.

“Understand the process. Design the system. Automate what matters.”

AI Automation
Workflow Architecture
Systems Integration
Backend Engineering
Data Architecture
Cloud Deployment
02 · Services

Put your tools to work together

I build the connections and tools that keep your business moving, so your team can spend less time on repetitive work.

Workflow automation

Keep information moving between your tools without copying it by hand. I build workflows for data entry, follow-ups, and reporting around the way your team works.

Business bots and messaging

Give your team and customers a simpler way to get things done. I build Telegram bots and WhatsApp integrations for common questions, record updates, and everyday requests.

Dashboards and internal tools

See the numbers that matter and act on them. I build custom dashboards and internal tools to track activity, manage records, and give your team a clearer view of the business.

Have a process that takes more time than it should? Let’s talk.

Discuss your project
03 · Selected work

From problem to product

A closer look at three projects: the problems behind them, the decisions that shaped them, and the results.

Project 01

Materiales San Rafael

Active · Client Project

A mobile credit and collections workflow for a family-owned building materials business.

AI Automation DeveloperBackend & DataSystems Integration

The context

Where the project started

The challenge

Customer credit and accounts receivable were tracked through handwritten notes, making balances difficult to update, verify, and audit.

A mobile-first workflow automation system for a family-owned building materials business. It digitizes customer credit, payments, balances, and account history while keeping the financial logic centralized and auditable.

The approach

How I worked through it

  1. 01I worked with the team to understand credit and collections, keeping the first release focused on their daily needs.
  2. 02I designed the customer, transaction, payment, and audit data before building the automation.
  3. 03I introduced the Telegram and n8n workflows in stages, checking them against real account activity.

The build

Key engineering decisions

  • ▸Supabase and PostgreSQL as the financial source of truth
  • ▸Self-hosted n8n workflows for business logic, validation, and integrations
  • ▸Telegram Bot API as a mobile operational interface for the team

Results

  • Automated 4 core accounts-receivable processes: customer registration, credit entries, payments, and balance inquiries
  • Replaced manual balance calculations with consistent database-driven updates
  • Validated financial logic through 5 charge, payment, and cancellation scenarios

What this demonstrates

I can turn a real business workflow into a dependable system, from the data model and business rules to a mobile interface the team can use every day.

All projects View repository

Project 02

PitchPilot AI

Project Complete

A voice-based AI pitch coach with live practice sessions and personalized feedback.

Full-Stack DeveloperBackend EngineerCo-founder

Inside the product

Live pitch practice with feedback you can use

The session evaluates the pitch in real time while the user maintains eye contact.
The system detects distractions during practice and includes them in the final feedback.
Each session ends with metrics, findings, and concrete actions for the next pitch.

The context

Where the project started

The challenge

People preparing a pitch often have to practice on their own, without immediate feedback from the kind of audience they will face.

A voice-based pitch practice platform powered by Gemini Live. Users rehearse with an AI playing an investor, teacher, or client, then receive feedback tailored to the session.

The approach

How I worked through it

  1. 01I mapped the experience into preparation, pitch, questions, coaching, and reporting, with explicit session states.
  2. 02I built a WebSocket proxy to coordinate live audio with Gemini while keeping API credentials on the server.
  3. 03I tested complete practice sessions and tuned voice detection and monitoring to reduce delays and interruptions.

The build

Key engineering decisions

  • ▸A Node.js and TypeScript backend that streams audio between the browser and Gemini Live over WebSockets
  • ▸Backend hosted on Google Cloud Run, with the web app deployed on Vercel
  • ▸Session monitoring that detects AI failures and helps prevent interruptions during practice

Results

  • Reduced perceived latency between turns by approximately 35% by optimizing voice activity detection and audio streaming
  • Reduced session interruptions from 4 to 0 during a two-day test period with 60 users
  • Generated session reports with performance feedback and practical next steps

What this demonstrates

I can take a live AI experience from prototype to deployment, bringing together audio streaming, session logic, and cloud infrastructure.

All projects View repository

Project 03

Zaaby App

Advanced MVP · TestFlight

An AI-powered iOS app that helps people turn saved links and ideas into useful next steps.

Full-Stack DeveloperBackend ArchitectCo-founder

Inside the product

Turn saved content into your next step

An organized library with filters, progress states, and direct access to suggested actions.
AI-generated summaries, tags, and next steps turn saved content into something actionable.
A progress dashboard tracks useful content, completed actions, and performance by category.

The context

Where the project started

The challenge

People save valuable content from social media but rarely turn it into action.

An AI-powered iOS productivity app that transforms saved links and personal ideas into summaries, structured takeaways, and actionable next steps.

The approach

How I worked through it

  1. 01I designed the journey from saving a link to receiving a summary, tags, and five personalized action items.
  2. 02I separated the mobile interface from server-side AI processing, with authentication and access rules around user data.
  3. 03I used TestFlight feedback to improve navigation, progress tracking, subscriptions, and AI usage costs.

The build

Key engineering decisions

  • ▸React Native, Expo, and TypeScript for the iOS application
  • ▸Supabase with PostgreSQL, Row Level Security, Auth, Storage, and Edge Functions
  • ▸OpenAI GPT-4.1-mini pipeline for summaries, tags, and action items

Results

  • Built approximately 80% of the product from architecture through beta deployment
  • Turned unstructured content into consistent, actionable AI output
  • Coordinated TestFlight beta testing with 10 users
04 · Approach

How I work

01

Understand before automating

I map the process, data, rules, and potential failure points before choosing the tools.

02

Systems over patches

If a problem repeats, I build a reliable system to prevent it from happening again.

03

AI where it adds value

Not every process needs AI. I use it when it improves decisions, speed, or the user experience.

04

Reliable by design

Validation, traceability, security, and error handling are part of the architecture from the start.

05

Build, measure, improve

I deliver working solutions, observe how they are used, and improve them with real-world results.

06

End-to-end ownership

I take projects from process analysis and architecture through integration, deployment, and continuous improvement.

05 · Skills

Tools I build with

AI & Automation

  • n8n · Workflow Automation
  • OpenAI API · GPT-4.1-mini
  • Google Gemini Live API
  • Telegram Bot API
  • Prompt & Output Design
  • AI Systems Integration

Backend & Integration

  • Node.js · TypeScript
  • Python
  • REST APIs · WebSockets
  • OAuth 2.0 · JWT
  • OpenAPI
  • Systems Integration

Data & Cloud

  • PostgreSQL · Supabase
  • Data Modeling · RLS
  • Edge Functions · Deno
  • Google Cloud Run
  • Docker · Containers
  • CI/CD · Vercel

Frontend & Mobile

  • React · Next.js
  • React Native · Expo
  • JavaScript · TypeScript
  • EAS Build · TestFlight

Technologies & platforms

n8n
Node.js
TypeScript
Python
Supabase
PostgreSQL
OpenAI
Gemini
Google Cloud
Docker
React Native
Expo
GitHub
Vercel
RevenueCat
n8n
Node.js
TypeScript
Python
Supabase
PostgreSQL
OpenAI
Gemini
Google Cloud
Docker
React Native
Expo
GitHub
Vercel
RevenueCat

“I use AI to move faster while staying accountable for the result. Understanding the system, checking its behavior, and making sound engineering decisions are still my responsibility.”

06 · Contact

Let’s make more room for the work that matters.

If your team is spending too much time copying data, chasing updates, or working around disconnected tools, I can help you build a better workflow.

Available for client projects and remote roles in AI automation and backend development.

Start a ConversationDownload CV
Donato Hernández© 2026 Donato Hernández. Building reliable systems and automations that solve real problems.