Master’s degree
2021 — 2023
GPA 4.00
Master of Computer Science and Information Engineering
National Central University · Taiwan
Specialization · Deep Learning
Available for AI engineering opportunities
I build AI that
works in the real world.
Hi, I’m Pink — an AI & Software Engineer turning complex LLM ideas into useful, reliable products.
pink.ai
class Pink(Engineer):
focus = [
"agentic AI",
"production RAG",
"reliable systems"
]
def build(self, idea):
return idea.to_product()
LLMs + thoughtful engineering ✦
01 / About
Curious mind.
Production mindset.
Pink · AI & Software Engineer
I’m Isariya Sirivejabandhu — call me Pink. I bridge AI experimentation and dependable software engineering.
My background spans full-stack development, cloud infrastructure, DevOps, and applied machine learning. Today, I focus on LLM applications: retrieval pipelines, agent workflows, evaluation, observability, and the product details that make AI genuinely useful.
4+Years in software
M.Sc.Computer Science
2Cloud certifications
Education02 degrees
Bachelor’s degree
2014 — 2018
Bachelor of Science in Computer Science
Kasetsart University · Thailand
02 / Experience
Where I’ve made
things happen.
From customer-facing platforms to cloud automation and AI systems.
2024 — 2025
Software Engineer
Ticketmaster Taiwan · Based in Thailand
- Worked as part of the Ticketmaster Taiwan engineering team while based in Thailand, developing and maintaining frontend and backend features for platforms across Singapore, Thailand, and the Philippines using PHP, Yii, JavaScript, HTML, CSS, Bootstrap, SQL, and AWS.
- Led engineering workflow transitions from Redmine to Asana and Jira, including board configuration, workflow rules, automation, Agile processes, and Confluence documentation.
- Designed and implemented Slack bot integrations for Jira using Node.js, AWS Lambda, and API Gateway, with automated deployment through GitLab CI/CD and Terraform.
- Supported infrastructure migration to Amazon EKS Auto Mode and managed internal services using Kubernetes, Helm, Argo CD, and GitOps practices.
2020 — 2021
Software Engineer
RS Public Company Limited
- Developed and maintained websites for record labels and entertainment brands, including Kamikaze, using custom themes and SEO best practices.
- Resolved production issues and implemented website enhancements to ensure system stability and on-time delivery.
- Collaborated with designers, content teams, and developers to deliver engaging, user-friendly web experiences.
2019 — 2020
Software Engineer
Pronto Marketing
- Developed responsive WordPress websites using HTML, CSS, JavaScript, PHP, and Bootstrap.
- Worked with cross-functional teams in an Agile environment to deliver and maintain customer web solutions.
- Provided technical support, resolved website issues, and mentored new hires on web technologies and development practices.
03 / Skills
Technical skills.
Technologies I have worked with across AI, software development, and cloud infrastructure.
AI & Machine Learning
PythonLarge Language ModelsLangChainLangGraph
LangSmithRAGQdrantMCPOpenAI API TensorFlowPyTorchNumPyscikit-learnpandas MatplotlibDeep Learning
Backend & Data
FastAPINode.jsPHPYii REST APIsPostgreSQLDynamoDBSQLNoSQL
Frontend & Web
JavaScriptTypeScript HTMLCSS / SCSSBootstrapWordPress
Cloud & DevOps
AWSDockerKubernetesTerraform HelmArgo CDGitGitHub ActionsCI/CD
Testing
PytestCodeceptionJest
Collaboration & Project Tools
JiraConfluenceSlackFigmaTrello AsanaRedmine
Languages
Thai Native
English TOEIC 805
Chinese Basic
Featured build2026
Enterprise AI
AI Document
Workspace.
Upload PDFs, organize them by conversation, and ask questions in natural language. The workspace finds relevant information, generates grounded answers, and shows the source pages used.
Open live app ↗
What it can do
Your research workspace for PDFs.
It brings document reading, source-backed Q&A, conversation history, and academic discovery into one workspace. Users can work across several PDFs without losing the context or source of an answer.
01
Upload and manage PDFsAdd multiple documents to separate conversation workspaces with progress, metadata, and duplicate-file feedback.
02
Ask document-grounded questionsChat naturally with one PDF or search across all documents in the active conversation.
03
Verify answers with citationsSee reranked sources and page-level references supporting each generated answer.
04
Keep research organizedCreate and rename conversations while preserving documents and complete chat history.
05
Discover academic papersSearch trusted academic sources and receive clickable, verified publication links.
06
Control sensitive actionsReview and approve or reject document deletion before the system makes the change.
Built with
Python · FastAPI · LangChain · LangGraph · LangSmith · OpenAI · Qdrant · PostgreSQL · Docker
05 / Selected web projects
Built before AI.
Still part of my story.
A selection of customer-facing websites delivered during my software engineering career.
Visit site ↗
Yasrebi Law
Responsive business website
Visit site ↗
Avenue Dental Plan
Healthcare membership website
Visit site ↗
Kamikaze Music
Entertainment and artist website
Archived project
RS Friends
Corporate community platform · Live site unavailable
Visit site ↗
Autocut
Automotive services website
Visit site ↗
Simplex
Technology services website
Additional work includes Cactus, FirstTrust Home Loans, JabberComm, Protec, RSiam, and other production websites.
06 / Research
Research that
connects the dots.
One research topic developed from my master’s thesis into a peer-reviewed Springer publication.
01
Master’s thesisOriginal research
Developed into
02
PublicationPeer-reviewed paper
Master’s thesis · National Central University
A Graph-based Approach for PM2.5 Prediction
Supervisor · Professor Min-Te Sun
Research role
Primary researcher
Responsible for the end-to-end research and experimental work.
The research proposed an end-to-end PM2.5 forecasting system covering data preprocessing, data fusion, feature engineering, feature selection, and multistep prediction.
I led the research design, data preparation, model development, experiment execution, evaluation, and analysis. The DCRNN-GS model combined DCRNN for spatio-temporal dependencies with GraphSAGE for learning meaningful information across sensor nodes. It achieved a 5.66% MSE improvement over the compared state-of-the-art approaches for forecasting the next 24 hours from the previous 24 hours.
A Graph-based System for PM2.5 Prediction
I am the second author of this peer-reviewed publication and was the primary contributor responsible for the research and experimental work. The paper presents the complete graph-based system and hybrid DCRNN–GraphSAGE model for 24-hour multistep air-quality forecasting.
Read the publication ↗07 / Certifications
Validated cloud
& platform skills.
Industry credentials supporting my ability to build and operate reliable AI systems.
AWS
Amazon Web Services
AWS Certified Solutions Architect — Associate
Cloud architectureSecurityScalability
Verify credential ↗
CKA
The Linux Foundation · CNCF
Certified Kubernetes Administrator
KubernetesNetworkingTroubleshooting
Verify credential ↗
Have a role or idea in mind?
Let’s build something
intelligent.
Isariyas213@gmail.com ↗
Based in Bangkok, Thailand · Open to global opportunities