DEV Community: Samuel Ferreira da Costa

Using Google AI Studio and Google Cloud for my 2k26 Porfolio

Samuel Ferreira da Costa — Mon, 02 Feb 2026 02:52:19 +0000

This is a submission for the New Year, New You Portfolio Challenge Presented by Google AI

About Me

I am a Brazilian software engineer who this year decided to independently launch four SaaS companies simultaneously:

Portfolio

https://cow-cow-ai-agency-portfolio-221884890346.us-west1.run.app

--labels dev-tutorial=devnewyear2026

MediStream: Patient triage with queue management, scheduling, and protection of sensitive patient data with humanized treatment; queue control and management of the anamnesis and examination flow.

OmniSell: Autonomous salesperson configured to handle everything from initial customer service via AI agents to closing the sale, with report generation and support for the company's internal marketing.

InsurFlow: Management of deadlines and workflows, encompassing everything from management to meeting deadlines for insurance companies.

TerraGuard: Works in land regularization with the aid of AI to verify documents required by Brazilian laws for land and plot regularization, helping producers and industries avoid unnecessary fees and fines.

How I Built It

Using Google AI Studio, I automated the creation of this portfolio, showcasing my products with high quality, and also allowed the AI to answer questions that potential clients might have about the problems.

I used simple, concise prompts, focusing on code simplicity. Enabling deployment to Google Cloud directly from AI Studio made my life much easier. I alternated between English and Portuguese to test the system's ability to understand what I was saying and was surprised. The project was based on JS/TS with a Node server.

Prompt 01:
I want to create a portfolio for an artificial intelligence engineering agency called COW COW. The agency's products are based on AI and currently include: a patient queue management system; a sales and product management system; a service and process management system for insurance companies; and an agricultural management and regulation system. For each of these systems, I want a navbar at the top to contain a button for each project. When clicked, the background, images, and colors should change to reflect the project's aesthetic. COWCOW excels in AI-powered management projects, and its portfolio should reflect this. Use appropriate colors for each product, and show that the company is ready to handle and receive projects for the next quarter.

Prompt 02:

User
quero que exista um switch para trocar o idioma da página, alternar entre ingles e portugues
Gemini 3 Flash Preview
Ran for 9s
lightbulb
Thought for 7 seconds
Para implementar o suporte multi-idioma (Inglês e Português) no portfólio da COW COW, faremos o seguinte:
Atualização de Tipos: Definiremos o tipo Language e estruturaremos os campos de texto nos projetos para suportar ambos os idiomas.
Reestruturação de Dados: Atualizaremos as constantes com as traduções para PT e EN.
Gerenciamento de Estado: Adicionaremos o estado de idioma no App.tsx.
UI do Switcher: Incluiremos um seletor elegante no Navbar.
Contexto para IA: Atualizaremos o serviço do Gemini para responder no idioma selecionado.

What I'm Most Proud Of

Personally, I'm using Antigravity (I retired VS Code) thanks to its native autocomplete and manager, which I use cautiously as my personal assistant, but which has served me very well. I can only thank you for the opportunity to participate in the challenge.

Communicate, Colaborate and Connect: A Holistic View

Samuel Ferreira da Costa — Mon, 28 Jul 2025 02:26:22 +0000

This is a submission for Frontend Challenge: Office Edition sponsored by Axero, Holistic Webdev: Office Space

What I Built

These tree words has be a summary of the objective that axero wanted in this challenge. So, I isolated the word in my view of a healthy landing page of the Andaime Inc, the company that thinks upwards.

Demo

See here the open code of project:(https://stackblitz.com/~/github.com/kailera/office-space)

Journey

In my learning journey, I decided to deepen my frontend knowledge with this challenge, using Next JS with Tailwind (frameworks are welcome) to structure a clean page that prioritizes employee well-being. The result:

However, employees may need to access the company website at night, perhaps even via their cell phone. So the solution:

Floating components minimize screen clutter while still enabling exactly what the challenge required: communication, collaboration, and connectivity. Feel free to access the project.

Tillage fields Analysis and Alerts with Runner H

Samuel Ferreira da Costa — Mon, 07 Jul 2025 01:05:37 +0000

This is a submission for the Runner H "AI Agent Prompting" Challenge

What I Built

I used Runner H and its tools to build a crop development analysis workflow.

Demo

Task 1: NDVI Analysis and Report Generate
https://www.loom.com/share/b3ad0c2a5af5455fa95df2ad9af2fe26?sid=88a74159-816a-433f-80e7-32dbc005981f

Task 2: Irrigation Recommendation By Google Sheets With Google Calendar
https://www.loom.com/share/af902ce8534242b99aefacdb9539c910?sid=9b9b3a95-d2d4-4ead-9b77-4f4f948cb9ff

Task 3: Florestal Fire With Gmail Alerts
https://www.loom.com/share/fad997c85ca942feb2e3b2012db2fb80?sid=fd9f6435-bd4d-4110-a946-763ce2bfca13

How I Used Runner H

Task 1: Weather data collected by an API made in FastAPI collects weather data from the OpenWeather API; Historical NDVI data from the agromonitoring.com API feeding a google Spreadsheet.
Using the Google Sheets connection provided by Runner H, I wrote the following prompt:
_using sheet NDVI in spreadsheet id: {secret}. For each different polygon Id, analyze the data, searching and given the informations below:
Technical Field Report Structure (Generated by AI)
🧾 1. Field Identification
Plot code or name

Location (GPS coordinates or QR/ID)

Cultivated area (m² or ha)

Crop(s) present

Soil type (clay, sandy, mixed)

Crop growth stage (planting, sprouting, flowering, harvest)

🌱 2. Crop Vegetative Status

Average NDVI and spatial variation (vegetation vigor)

Crop health classification (good, moderate, critical)

Detection of gaps, planting failures, or water stress

Biomass and canopy coverage estimate

Thermal and water stress indices (ET0, water deficit)

Estimated Productivity Harvest forecast based on history, climate, and NDVI

Comparison with previous cycles or similar plots

Margin of error based on climate uncertainty

Suggested optimal harvest window

AI-Based Technical Recommendations
Based on previous sections:

Priority agronomic interventions (irrigation, fertilization, pesticide)

Fertilization suggestions based on soil + crop stage

Soil management (subsoiling, cover cropping, intercropping)

Alerts for harvesting, replanting, or operational failures

📄 8. Executive Summary
Overall plot condition (agronomic performance index)

Key risks and opportunities

Graphs: NDVI, climate, yield, heatmaps

create a file in @tool:Google docs and transcribe all the analysis_

Task 2: Using weather data provided by the API and the connection to Google Spreadsheet, I wrote a prompt that analyzed the weather and gave recommendations for irrigation in the registered crops. I asked for links to be generated per day to be inserted into Google Calendar. The prompt was:
_Access the @tool:Google sheets the sheet FORECAST in file id {secret}
Analyze the probability of rain and the air humidity, and estimate the irrigation need in fields. data provided: Temp Day Temp Min Temp Max Temp Night Temp Pressure Humidity Weather Description Prob. Rain Rain (mm) For each day,

Schedule in @tool:Google sheets the file with id {secret} in the sheet Event Link
the irrigation program for 10 days, separated by links to insert in google calendar days with the information:

Title of the Event: Irrigation Info
Date and Time: all day
Location: city or location of the forecast
Description: irrigation orientation (recommendation or not) and other pertinent infos

each link must be inserted in the direction of the rows. Just the links must be inserted in sheet. Don't need to create a pdf._

Task 3: For this task, I used the connection to Google Sheets and Gmail to access the latest updated records of large fires in Brazil and calculate which ones were close to the monitored crops. If there was one nearby, Runner H would send an email alerting about imminent danger. Since there wasn't, a monitoring message was sent via Gmail. The prompt was:
_Access the @tool:Google sheets the sheet POLYGONS in file id {secret}
and memorize the locations indicated by latitude and longitude. Each row represents one location as a farm or field,
demarcated by latitude and longitude.

Access the @tool:Google sheets the sheet {secret}
Analyze the locations indicated by latitude and longitude of florestal fires. If there is anyone closer than 300 kilometers
of one farm or field memorized, use the @tool:Gmail and send an email to {secret} with title: Florestal Fire Infos and in Subject writes if there are florestal fires near. If don't have forest fires near,writes in subject: No Florestal Fires near in your location_

Use Case & Impact

Family farming produces around 70% of the food consumed in Brazil. However, extreme weather events have posed a major risk to these producers and agronomists who work hard to overcome these challenges. Using an extremely advanced AI system like Runner H, much of the bureaucratic work of consolidating crop data provided by satellites, climate data and APIs can be automated, accelerating decision-making that needs to be accurate and fast for crop growth or even protection.

Social Love

My GitHub: github.com/kailera