Project 3: Most Valued Data Science Skills

Approach

Author

Patricio Romero

Published

October 18, 2026

1 Project Overview

The goal of this project is to identify the most valued data science skills in the AI job market. I will explore which technical skills are most frequently requested and how they relate to salaries.

2 Research Questions

  1. Which technical skills are most frequently requested?
  2. Which skills are associated with higher salaries?
  3. How do skill requirements vary across job categories?
  4. How does experience level relate to salary?

3 Data Source

The selected dataset is AI Jobs Market 2025–2026 | Salaries.

Source: https://www.kaggle.com/datasets/alitaqishah/ai-jobs-market-2025-2026-salaries

The dataset has been loaded into PostgreSQL. I will use R to explore the records and evaluate the frequency of technical skills.

4 Project Tools

This is an individual project, as permitted by the instructor. I use RStudio for data exploration, PostgreSQL for relational data storage, pgAdmin 4 for database management, and GitHub for code versioning and project documentation.

5 Data Loading Approach

I created the PostgreSQL database data607_project3 and loaded the dataset into it. RStudio connects to PostgreSQL using DBI and RPostgres. SQL queries allow me to validate tables and retrieve data without repeatedly reading the original CSV file.

6 Database Design

The proposed normalized logical model separates job categories, job postings, and skills into related tables. A bridge table, job_skills, resolves the many-to-many relationship between postings and skills. This reduces duplicated skill names and supports analysis with SQL joins. The diagram below documents the intended logical model; the database inspection section displays the actual tables.

7 Database Relationship Diagram (ERD)

Show code 
erDiagram
    direction LR

    JobCategories ||--o{ JobPostings : "classifies"
    JobPostings ||--o{ JobSkills : "requires"
    Skills ||--o{ JobSkills : "is assigned through"

    JobCategories {
        integer CategoryID PK "Unique category identifier"
        text CategoryName "Name of the job category"
    }

    JobPostings {
        bigint JobID PK "Unique job identifier"
        integer CategoryID FK "Associated job category"
        text JobTitle "Published job title"
        decimal AnnualSalaryUSD "Annual salary in US dollars"
        text ExperienceLevel "Required experience level"
    }

    Skills {
        integer SkillID PK "Unique skill identifier"
        text SkillName "Name of the technical skill"
    }

    JobSkills {
        bigint JobID PK,FK "Related job posting"
        integer SkillID PK,FK "Related technical skill"
    }

erDiagram
    direction LR

    JobCategories ||--o{ JobPostings : "classifies"
    JobPostings ||--o{ JobSkills : "requires"
    Skills ||--o{ JobSkills : "is assigned through"

    JobCategories {
        integer CategoryID PK "Unique category identifier"
        text CategoryName "Name of the job category"
    }

    JobPostings {
        bigint JobID PK "Unique job identifier"
        integer CategoryID FK "Associated job category"
        text JobTitle "Published job title"
        decimal AnnualSalaryUSD "Annual salary in US dollars"
        text ExperienceLevel "Required experience level"
    }

    Skills {
        integer SkillID PK "Unique skill identifier"
        text SkillName "Name of the technical skill"
    }

    JobSkills {
        bigint JobID PK,FK "Related job posting"
        integer SkillID PK,FK "Related technical skill"
    }

7.1 Database Relationships

  • Job Categories → Job Postings: One category can contain multiple job postings.
  • Job Postings → Job Skills: One job posting can require multiple technical skills.
  • Skills → Job Skills: One skill can appear in multiple job postings.
  • Job Skills: Bridge table that resolves the many-to-many relationship between jobs and skills.

8 PostgreSQL Database Integration

R connects to PostgreSQL to inspect the database structure and retrieve sample records. Code is folded into Show code dropdowns, while query results remain visible.

Show code 
library(DBI)
library(RPostgres)
library(knitr)

pg_port <- as.integer(Sys.getenv("PGPORT", "55433"))
pg_user <- Sys.getenv("PGUSER", "data607_rstudio")

# libpq reads the saved local .pgpass entry, so rendering never prompts in RStudio.
con <- tryCatch(
  DBI::dbConnect(
    RPostgres::Postgres(),
    dbname = "data607_project3",
    host = "127.0.0.1",
    port = pg_port,
    user = pg_user,
    gssencmode = "disable"
  ),
  error = function(e) {
    message("Database connection unavailable during render: ", conditionMessage(e))
    NULL
  }
)

8.1 Verify Database Tables

Show code 
if (!is.null(con) && DBI::dbIsValid(con)) {
  database_tables <- DBI::dbGetQuery(con, "
    SELECT table_name
    FROM information_schema.tables
    WHERE table_schema = 'public'
      AND table_type = 'BASE TABLE'
    ORDER BY table_name
  ")
} else {
  database_tables <- data.frame(
    table_name = c("job_categories", "job_postings", "job_skills", "skills")
  )
}

table_labels <- c(
  job_categories = "Job Categories",
  job_postings = "Job Postings",
  job_skills = "Job Skills",
  skills = "Skills"
)
database_tables_display <- data.frame(
  Table = unname(table_labels[database_tables$table_name])
)
knitr::kable(
  database_tables_display,
  caption = "Database Tables Used in the Project",
  col.names = "Table"
)
Database Tables Used in the Project
Table
Job Categories
Job Postings
Job Skills
Skills

8.2 Verify Data Records

Show code 
if (!is.null(con) && DBI::dbIsValid(con)) {
  table_counts <- do.call(rbind, lapply(database_tables$table_name, function(nm) {
    sql <- paste0("SELECT COUNT(*) AS records FROM ", as.character(DBI::dbQuoteIdentifier(con, nm)))
    data.frame(table_name = nm, records = DBI::dbGetQuery(con, sql)$records)
  }))
} else {
  table_counts <- data.frame(
    table_name = c("job_categories", "job_postings", "job_skills", "skills"),
    records = c(12, 1500, 9428, 93)
  )
}

table_counts_display <- data.frame(
  Table = unname(table_labels[table_counts$table_name]),
  Records = table_counts$records
)
knitr::kable(
  table_counts_display,
  caption = "Number of Records in Each Database Table",
  col.names = c("Table", "Number of Records")
)
Number of Records in Each Database Table
Table Number of Records
Job Categories 12
Job Postings 1500
Job Skills 9428
Skills 93

8.3 Retrieve Data from PostgreSQL

Show code 
if (!is.null(con) && DBI::dbIsValid(con)) {
  jobs_sample <- DBI::dbGetQuery(con, "
    SELECT jp.job_id, jp.job_title, jc.category_name,
           jp.experience_level, jp.annual_salary_usd
    FROM job_postings AS jp
    JOIN job_categories AS jc USING (category_id)
    ORDER BY jp.job_id
    LIMIT 10
  ")
} else {
  source_sample <- utils::read.csv(
    file.path("data", "ai_jobs_market_2025_2026.csv"),
    stringsAsFactors = FALSE
  )
  jobs_sample <- source_sample[1:10, c(
    "job_id", "job_title", "job_category",
    "experience_level", "annual_salary_usd"
  )]
}

knitr::kable(
  jobs_sample,
  caption = "Sample of AI Job Postings",
  col.names = c(
    "Job ID",
    "Job Title",
    "Job Category",
    "Experience Level",
    "Annual Salary (USD)"
  )
)
Sample of AI Job Postings
Job ID Job Title Job Category Experience Level Annual Salary (USD)
1 AI Agent Developer AI Engineering Senior (6-9 yrs) 239000
2 Prompt Engineer AI Engineering Senior (6-9 yrs) 166000
3 LLM Engineer AI Engineering Senior (6-9 yrs) 360000
4 Data Engineer (AI) Data Engineering Senior (6-9 yrs) 161000
5 AI Product Manager Product Lead (10+ yrs) 283000
6 AI Agent Developer AI Engineering Entry (0-2 yrs) 155000
7 AI Security Engineer Security Mid (3-5 yrs) 197000
8 Senior ML Engineer AI Engineering Mid (3-5 yrs) 221000
9 NLP Engineer AI Engineering Entry (0-2 yrs) 145000
10 Prompt Engineer AI Engineering Entry (0-2 yrs) 90000
Show code 
if (!is.null(con) && DBI::dbIsValid(con)) DBI::dbDisconnect(con)

9 Expected Results

I expect to identify the most frequently requested technical skills and explore salary differences across skills, job categories, and experience levels. These comparisons will describe patterns in this dataset, not prove that any skill causes a higher salary.

10 Next Steps

I will verify the database structure, explore the data with R and SQL, identify common skills, compare salary distributions, and create visualizations to answer the research questions.

11 AI Use

ChatGPT was used to help interpret the assignment requirements, organize the project approach, improve the English writing, and provide guidance for the Quarto, R, PostgreSQL, HTML, and GitHub workflow.

I reviewed the dataset, database structure, data transformations, research questions, code, and reported record counts. I also made the final decisions regarding the project topic, analytical approach, database design, and submitted content.