AI automation engineer and full-stack developer.

I build AI-powered automation workflows, LLM integrations, and full-stack web applications, from the backend to the deployed product.

See the work Email me GitHub
status_readout open to roles
location
Metro Manila, PH
local time
—:—:—
focus
ML / automation / full-stack
Portrait of Christian Joshua Alberto
Your photoimages/cj-portrait.jpg

I'm Christian Joshua Alberto, a Computer Science graduate (Magna Cum Laude) from National University – Manila with a specialization in machine learning. My work focuses on building software that runs reliably in production, including machine learning models, automated data pipelines, and full-stack web applications.

Recent projects include a dengue outbreak forecasting model deployed with a live API and public dashboard, an automated market analysis pipeline that delivers daily reports after each Philippine Stock Exchange close, and an end-to-end test suite that runs on every pull request. During my internship, I automated HR workflows that cut manual processing time by 80%. I'm currently looking for roles in AI/ML, backend, or full-stack development.

I focus on building systems that are dependable, maintainable, and ready for real use.

Selected work

Twelve builds, sorted by what they're made of. Filter to narrow.

year2026
typeAI resume screener
stackn8nGoogle Gemini APIJavaScriptHTML/CSSGitHub Pages

Resume Fit Analyzer

AI resume screener that scores how well a resume fits a job description

Paste a job description and upload a resume PDF to get a fit score out of 100, sub-scores for skills, experience, and education, matched and missing skills, a summary, and a hiring recommendation. It works for any industry, reads scanned PDFs, and caps the score when a must-have requirement is missing. It pairs with Interview Prep Coach: after an analysis, a “Practice the gaps” button opens the coach with the job description and missing skills carried over. Both use Gemini with structured JSON output and fall back to a lighter model automatically if the primary one fails, and a “Try a sample” button lets you test it without your own resume.

100
point fit score, capped when a must-have is missing
3
sub-scores: skills, experience, education
Try it live View source
Resume Fit Analyzer showing a 66 out of 100 Partial Fit score with skills, experience, and education sub-scores, a summary, and matched skills
Fit score, sub-scores, and matched skills/images/projects/resume-fit-analyzer.png
year2026
typeAI interview practice
stackn8nGoogle Gemini APIJavaScriptHTML/CSSGitHub Pages

Interview Prep Coach

AI interview practice tailored to a specific job and resume

Generates 6 likely interview questions from a job description and resume, including questions that probe gaps. Each typed answer gets a score out of 10 on a four-part rubric, specific fixes, a stronger version written only from the candidate's real experience, and a follow-up question. It runs multi-turn sessions without a database by keeping state in the browser. It picks up where Resume Fit Analyzer leaves off: its “Practice the gaps” button opens the coach with the job description and missing skills carried over. Both use Gemini with structured JSON output and fall back to a lighter model automatically if the primary one fails, and a “Try a sample” button lets you test it without your own resume.

6
tailored questions per session, including gap probes
4
rubric criteria scored on every answer
Try it live View source
Interview Prep Coach scoring a behavioral answer 7 out of 10 with rubric ratings, what worked, what to improve, and a stronger version of the answer
Answer score, rubric, and stronger version/images/projects/interview-prep-coach.png
year2026
typeLLM app · live demo
stackCloudflare WorkersGeminiJavaScriptVercel

Portfolio AI Assistant

The chat on this page — you can try it right now

A chatbot that answers questions about my work, running on Cloudflare Workers and streaming Gemini replies into a chat widget I built from scratch. It's grounded in a curated knowledge base of this portfolio and told to decline off-topic requests and admit when it doesn't know something rather than invent a detail. Because it's public, it's built like a public service: the API key never reaches the browser, only this site may call it, each visitor is rate-limited, and if the main model is overloaded a backup model takes over automatically.

2
models: primary plus automatic backup
0
API keys exposed to the browser
Source code
year2026
typeAutomation agent
stackn8nGeminiYahoo FinanceTelegramSupabase

PSE Daily Market Brief

A market-analysis agent that reports itself

An n8n pipeline that wakes up after every Philippine Stock Exchange close, pulls the index, a 13-stock watchlist and five macro series from Yahoo Finance, reads three news feeds, and sends a Gemini-written brief to Telegram — session summary, top movers, and sentiment tagged per headline. The hard part was keeping the model honest: links are stitched back from the source feeds in code so it can't invent URLs, output is validated, and a failed call falls back to a second model. Every run logs prices and sentiment to Postgres, with SQL views tying each day's sentiment to the next day's return so I can backtest whether the signal is real.

13
stocks + 5 macro series tracked daily
0
model-generated links (all sourced in code)
PSE Daily Market Brief delivered in Telegram
Screenshot of a brief in Telegramimages/pse-brief.webp
year2026
typeQA automation platform
stackFastAPIReactTypeScriptPlaywrightGitHub Actions

TaskFlow-QA

A task tracker built to be tested

A full-stack task tracker I built specifically as the system-under-test for a Playwright framework — E2E and API suites on a Page Object Model. Working with Claude Code I hunted down two genuinely nasty reliability bugs: an async race in a page-object helper and a shared-state leak across parallel workers. Fixing them took the suite from intermittent failures to a clean, repeatable pass across Chromium and Firefox. CI type-checks the frontend, runs backend tests, and executes the full cross-browser suite with an uploaded HTML report on every PR.

18 → 0
flaky failures eliminated
30/30
consistent pass, two browsers
TaskFlow-QA Playwright test report
Playwright report or green GitHub Actions runimages/taskflow-qa.png
year2026
typeProduction event platform
stackFastAPISupabaseThree.jsPayMongoGmail API

Exclusives PH — Yacht Club event

Freelance · full ticketing stack, live traffic

A full-stack RSVP and ticketing platform for a Manila Yacht Club event, built from scratch and run against real crowd traffic. PayMongo handles GCash and Maya payments, the Gmail API fires QR e-tickets the moment someone pays, and a reception dashboard scans those codes at the door and assigns table seating automatically. I also built the on-site point-of-sale and bar management systems to keep transactions moving during the event.

100%
of manual booking work removed
Live
handled event-night traffic end to end
Live site
Exclusives PH ticketing platform
Ticketing page, QR check-in, or event nightimages/exclusives-ph.webp
year2025–26
typeDeep learning thesis
stackTensorFlowGCPRenderVercel

Dengue Early-Warning System

My undergraduate thesis, deployed for real

A spatial-temporal deep learning model that forecasts dengue outbreaks across Metro Manila from clinical and weather data. Beyond the modelling — preprocessing, feature engineering, hyperparameter tuning — I shipped it to production behind a live REST API with a public dashboard, so the forecast is something you can actually open in a browser rather than a figure in a paper.

Live dashboard
Dengue early-warning dashboard
Screenshot of the live dengue dashboardimages/dengue-dashboard.png
year2026
typeLLM service
stackFastAPIAWS BedrockClaudeStructured JSON

AI Resume-Fit Analysis

The same tool, on a second cloud

A FastAPI service on Amazon Bedrock (Anthropic Claude) that scores a resume against a job description and returns matching skills, gaps, and a recruiter-style summary as structured JSON. I built it to mirror an OpenAI/n8n tool already running in production, so I could compare the two clouds' LLM integration head to head — latency, output reliability, and how much prompt work each needed.

AI resume-fit analysis output
API response or fit-score outputimages/resume-fit.png
year2025
typeDeep learning
stackTensorFlowSiamese NN

Signtinel

Signature forgery detection

A Siamese neural network that verifies signatures and flags high-precision forgeries, validated with Euclidean distance metrics and a full confusion-matrix analysis rather than a single accuracy number.

year2024
typeIoT + computer vision
stackArduinoOpenCVGSM

XORBIN — Smart Waste System

Developer lead · packetHACKS 2024 finalist (4th)

An IoT waste-classification system that sorts material with computer vision, driven by Arduino and ultrasonic sensors, with a GSM module for real-time alerts. I led the developer team through the build to a 4th-overall finish at the packetHACKS NCR leg.

XORBIN smart waste bin prototype
Photo of the XORBIN prototypeimages/xorbin.jpg
year2024
typeClassical ML
stackscikit-learnSVM

Heart-Failure Risk Prediction

Clinical risk classifier

A Support Vector Machine that predicts heart-failure risk from clinical data, with the full pipeline handled end to end — cleaning, feature scaling, and tuning — rather than a black-box fit.

year2024
typeReinforcement learning
stackDeep Q-LearningOpenAI Gym

Autonomous Taxi Navigation

Deep Q-Learning agent

A reinforcement-learning agent trained on OpenAI Gym's Taxi-v3 that learns pick-up and drop-off routing on its own, with no predefined map handed to it.

Experience & education

Where the reliability habit came from.

Jul – Aug 2026

Freelance Full-Stack Developer

Exclusives PH — Manila Yacht Club event

  • Architected the RSVP and ticketing platform end to end with FastAPI, Supabase, and an interactive Three.js frontend that held up under live event traffic.
  • Automated payments (PayMongo / GCash / Maya) and instant QR e-ticket dispatch via the Gmail API, removing manual booking entirely.
  • Built the door-side reception dashboard, seat assignment, and an on-site POS and bar management system.
Mar – May 2026

Automation Engineer Intern

Crowdsource Innovative Solutions Inc.

  • Cut manual HR processing time by 80% with n8n workflows over Google Workspace APIs, PostgreSQL, and Supabase.
  • Automated a 3-tier approval pipeline, timesheet updates, and document generation for a 16-person organisation.
  • Shipped a production HR portal with real-time dashboards and webhook integrations, plus an AI resume-fit analyzer using the OpenAI API and structured JSON outputs.
Aug 2022 – Sep 2026

BS Computer Science, Machine Learning

National University – Manila

  • Graduated Magna Cum Laude and was a Consistent First Honor Dean's Lister.
  • Thesis: a deep learning early-warning system for dengue in NCR, deployed with a live dashboard and REST API.
Aug 2024 – Nov 2024

Internal Vice President

Google Developer Student Clubs

  • Led operations across three committees and raised member engagement by 20% through technical workshops.
  • Cut event prep time by 25% by streamlining the club's internal workflows.

What I build with

Grouped by what I reach for it to do.

Machine learning & data

TensorFlowKerasscikit-learnPandasNumPyRegressionClassificationTime-seriesComputer visionReinforcement learning

Backend & web

PythonFastAPIREST APIsReactTypeScriptTailwind CSSThree.jsHTML5CSS3

Automation & LLMs

n8nOpenAI APIGemini APIPrompt engineeringStructured JSONTelegram Bot APIClaude Code

Cloud, data & testing

PostgreSQLSupabaseMongoDB AtlasGCPAWS BedrockCloudflare WorkersVercelRenderDockerPlaywrightGitHub Actions

Languages — Python · Java · JavaScript · SQL · C++

Recognition

Competitions & honors

20241st Runner-Up, Google Solution Challenge IdeathonPitcher & system-logic lead — SDGs 3, 9, 17
2024Finalist (4th overall), packetHACKS NCRDeveloper lead, XORBIN smart-waste system
2023Finalist (Top 8), DLSU Hacker CupSustainable-tech prototype designer
—Magna Cum Laude & Consistent First Honor Dean's ListerNational University – Manila

Certifications & results

2025TOPCIT PH — Level 3, 576/1000Top 10% nationwide, 12th sitting
—Google Data Analytics CertificateData cleaning, analysis, visualisation
—TOEICReading 460 · Listening 495