Data Specialist, Backend AI Intern, Published Researcher, AI Workflow Analyst.
I work at the point where data becomes a decision: dashboards and reports that people actually read, and the backend and automation underneath that keep them current without anyone re-typing a number.
Day to day I track a Scholarship program with status as Enrolled, Graduated, Assessed, Certified and Employed. Alongside that I'm interning as an AI backend engineer, and as a published first-author researcher I've run the full cycle of turning raw records into a conclusion that holds up when someone checks it.
"Best Student Research: CS, Data Science and Algorithms Category, NDMU 18th Araw ng Parangal"
],
"thesis": "Published @ IJLTEMAS"
}
Joshua Klyne Pudadera
$ cat about.md
I check things before they ship
Computer Science, Notre Dame of Marbel University, 2025. Through 2026 I ran an AI-assisted research and writing pipeline that produced long-form technical articles on AI and software engineering, with a verification gate between draft and publication.
Alongside that, nearly five years maintaining records for a TESDA-registered training institution to government documentation standards and audit checklists. That work is where the habit comes from: a record is either audit-ready or it is not, and no amount of confidence closes that gap.
I'm currently a Backend AI Engineer intern at FlyRank AI, working from written specifications and shipping written documentation with every deliverable.
# Banga, South Cotabato, Philippines · open to remote
4+Years of audit-ready records
2,918Rows reconciled, dengue dataset
3Vocational programmes covered
1Peer-reviewed paper, author out of four authors
/tools-used
Claude codeChatGPTPrompt EngineeringMicrosoft WordMicrosoft ExcelGoogle SheetsLooker Studio
Certifications
Oracle Cloud Infrastructure 2025 Certified AI Foundations Associate
Introduction to Machine Learning on AWS
AWS Educate Introduction to Generative AI
Leveraging AI for Text Analysis: NLP Using Orange Data Mining
Dashboards and analysis first, then the engineering that supports the data work and keeps it running. Written from what I have shipped, not what I have read about.
Git & GitHubGitHub Actions CI/CDCode review of AI outputTechnical documentationVersion-controlled revisionsVerification before mergeArchitecture & system designFull-stack delivery
$ cat stack.json
Tech stack
The tools the work actually runs on. Analysis on top, dashboards and documentation in the middle, what ships at the bottom.
Claude
Gemini
Hugging Face
Python
Jupyter
Colab
pandas
NumPy
scikit-learn
Streamlit
Claude
Gemini
Hugging Face
Python
Jupyter
Colab
pandas
NumPy
scikit-learn
Streamlit
Looker Studio
MySQL
GitHub Actions
QGIS
Google Earth
Sheets
Docs
Drive
Forms
Notion
Markdown
Looker Studio
MySQL
GitHub Actions
QGIS
Google Earth
Sheets
Docs
Drive
Forms
Notion
Markdown
JavaScript
TypeScript
React
Node.js
FastAPI
PostgreSQL
Git
GitHub
Flutter
Dart
Android Studio
Figma
Postman
JavaScript
TypeScript
React
Node.js
FastAPI
PostgreSQL
Git
GitHub
Flutter
Dart
Android Studio
Figma
Postman
$ ls ./projects
Projects
Eight projects, with what each one actually produced.
EGACE status dashboard
Dashboard
A recurring Excel dashboard tracking every scholarship trainee through the five EGACE stages — Enrolled, Graduated, Assessed, Certified, Employed — built on lookups against the trainee records so the counts update from one source rather than being tallied by hand.
→ One place to see where a batch stands, instead of five separate counts.
ExcelLookupsChartingStatus reporting
update-github-info.yml
on:
workflow_dispatch:
schedule: [cron: daily]
permissions:
contents: read
safe-outputs:
create-pull-request:
draft: true
Agentic CI/CD workflow
DevOps
A GitHub Skills challenge completed end to end: an agentic GitHub Actions workflow authored in markdown and compiled to a .lock.yml pipeline. Runs on a daily schedule or on demand with read-only permissions, fetches sources over an explicit network allowlist, keeps its credentials in repository secrets, and returns its change as a draft pull request instead of pushing to main.
→ Triggers, least privilege, secrets and a human review gate.
Two custom Claude Code skills that automate the branch-to-PR loop for teams using Linear with the GitHub integration. Linear links a branch to an issue only by matching the branch name it would have generated itself, so create-feature-branch reproduces that format exactly rather than re-implementing the slug, and create-pr runs the project's checks, summarizes the real diff, and updates an open PR instead of failing on one.
→ Read-only against Linear, so a two-way sync can't produce duplicate issues.
Claude Code skillsLinear MCPGitHub CLIGit workflow
A monthly 2015–2024 dataset for SOCCSKSARGEN, reconciled from DOH-CHD case records, Google Earth Engine rasters and NAMRIA shapefiles, and the forecasting model built on it.
→ 2,918 rows × 19 columns from three incompatible source types. First author of four.
A repeatable pipeline built in Claude: topic and niche validation, source gathering, automated scraping with Firecrawl, claim verification against official documentation and peer-reviewed sources, then an editorial pass for voice and consistency.
→ Ran on every article, so the process stayed auditable and each claim traceable.
ClaudeFirecrawlPrompt engineering
Skill output landing in Google Docs
medium-draft, a custom Claude skill
Tooling
A Claude skill that outputs publication-ready drafts directly into Google Docs, so research and publishing meet without a manual reformatting step in between.
→ Removed the reformatting step entirely.
Claude skillsGoogle DocsWorkflow design
Transaction tracking mobile app
Internship
Built at LEADSolutions INC as an on-the-job trainee: a Flutter mobile application for transaction tracking, integrated with the company's existing web system on Azure Cloud so data moved between the two without manual re-entry.
→ Designed the UI/UX as well as the build, from navigation through to cloud integration.
Trainee, programme and scholarship records encoded into T2MIS, TESDA’s national training management information system, plus the compliance documentation behind them and the billing reconciled against government purchase orders.
→ Three vocational programmes kept audit-ready for four years running.
T2MISComplianceBilling reconciliation
$ cat pipeline.md
Writing, and the gate before it
Long-form technical articles on AI and software engineering, 1,200 to 2,500 words, for professional audiences on Medium and LinkedIn. Every one of them went through these five steps.
01
Topic and niche validation
Is there an audience, and is there anything new to say to them.
02
Source gathering
Government, academic and technical documentation first, commentary second.
03
Automated scraping
Firecrawl pulls the sources so the reading list is complete before drafting.
04
Claim verification
Every technical claim checked against primary sources. Nothing publishes on trust.
05
Editorial pass
Voice and consistency, plus diagrams so verified facts read at a glance.
›Work from written specifications and ship written documentation with every deliverable.
›Review AI-generated output against my own work line by line and document the differences.
›That audit surfaced a logic error neither version flagged on its own.
SpecificationsDocumentationAI output review
Nov 2021 – present
Data Entry Specialist, Processing Officer / Registrar, Compliance Officer
Nenita Farm Rice-Based Farm Training Center
›Build the Excel dashboards and status reports that track each batch across the EGACE stages: Enrolled, Graduated, Assessed, Certified, Employed, so programme staff can see where a cohort stands without opening the underlying records.
›Encode trainee, programme and scholarship records into T2MIS and maintain the compliance documentation behind them.
›Keep records audit-ready across three vocational programmes, working to agency documentation standards rather than internal convenience.
›Coordinate scholarship batch records with TESDA offices through the full lifecycle, from enrolment through training and assessment to billing.
›Prepare billing statements against government purchase orders.
›Long-form technical articles on AI and software engineering for professional audiences.
›Built the research pipeline behind them in Claude.
›Generated explanatory diagrams and facts sheets so verified facts read at a glance.
›Shipped a custom skill that puts publication-ready drafts straight into Google Docs.
ClaudeFirecrawlMediumDiagramsApify
Jun 2024 – Jul 2024
On-the-Job Trainee
LEADSolutions INC, Koronadal City
›Developed a mobile application for transaction tracking in Flutter and Dart with Android Studio.
›Integrated it with the company's existing web system hosted on Azure Cloud for seamless data exchange.
›Designed the UI/UX so navigation and functionality stayed intuitive for the people entering transactions daily.
Flutter / DartAndroid StudioAzure CloudUI/UX
$ mail -s "hello" klynejoshua13@gmail.com
Interested in working together?
Open to full-time work in research, verification and technical writing, and to freelance or consulting engagements where claims need checking before they ship. Backend AI and documentation work included.