AI Engineer · Data EngineerPakon Poomson

AI Engineer & Data Engineer building reproducible data systems, ML workflows, and AI products.

I design local-first, CI-verified projects across data platforms, RAG systems, ML/MLOps, Thai NLP, and privacy-aware AI applications.

View Projects Resume GitHub LinkedIn

Customer support RAG triage dashboard and workflow.

Selected engineering evidence

Customer Support RAG Triage Agent

Demo available

Data systems: Local-first pipelines, analytical models, APIs
AI systems: RAG evaluation, guardrails, multimodal workflows
Verification: Deterministic demos, tests, CI, limitations

Featured Projects

Priority case studies across data engineering, RAG systems, and ML/MLOps, backed by public repository evidence.

Demo availableRAG / AI Agent Systems

Customer Support RAG Triage Agent

Offline-capable RAG triage system for support tickets with grounded responses.

Role signal: Retrieval, ranking, guardrails, evaluation, and API contracts.

What I built

System: A retrieval-grounded support triage workflow with observable steps, citations, evaluation, and fallback behavior.
Implementation: Typed workflow state keeps triage behavior inspectable.

What it proves

Skills: Retrieval, ranking, grounded generation, guardrails, offline evaluation, API contracts, and full-stack delivery.
Core stack: Python, FastAPI, RAG, LangGraph

RAGGroundingEvaluationGuardrails

Details GitHub Demo

Case studyRetail Data Engineering / Quality Platform

RetailGuard Data Platform

Zero-cost local retail data platform with incremental Bronze extraction, protected Silver data, blocking quality checks, and DuckDB warehouse evidence.

Role signal: Incremental pipelines, PySpark, DuckDB warehousing, quality gates, and privacy-aware data engineering.

Deterministic diagram of the RetailGuard local Bronze-to-Silver quality-gated DuckDB path.

What I built

System: A local retail analytics platform spanning source seeding, incremental extraction, privacy-aware transformation, quality gates, warehouse loading, and evidence reporting.
Implementation: Default review path is fully local and requires no cloud account, billing account, free trial, or hosted service.

What it proves

Skills: Data engineering, PySpark transformations, warehouse modeling, privacy controls, quality gates, idempotency, Docker, and local-first reviewer workflows.
Core stack: Python, PySpark, DuckDB, PostgreSQL

Data engineeringPySparkDuckDBQuality gates

Details GitHub

Demo availablePublic Data / Procurement Analytics

Thai Procurement Intelligence

Bilingual procurement intelligence platform using a governed 250-record official DGA/CGD snapshot.

Role signal: Public-data ingestion, provenance, bilingual evidence UI, and validation.

Official local snapshot evidence for the Thai procurement intelligence platform.

What I built

System: A bilingual public-data analytics workflow spanning official-source ingestion, SHA-256 verification, provenance, validation, evidence views, quality reporting, and source-linked assistant retrieval.
Implementation: The local official-snapshot mode uses a governed 250-record DGA/CGD snapshot with SHA-256 verification.

What it proves

Skills: Public-data ingestion, provenance design, bilingual UX, quality controls, security checks, and CI-safe analytics delivery.
Core stack: Python, FastAPI, Next.js, Data validation

Public dataProvenanceBilingualValidation

Details GitHub Demo

Explore all 8 case studiesData engineering, AI systems, ML/MLOps, Thai NLP, and public-data evidence

Engineering Evidence

Durable proof points used across the featured repositories, without relying on private infrastructure or undocumented claims.

Verification

Deterministic offline demos keep core reviewer flows reproducible.
CI, tests, data validation, and API contracts verify project boundaries where documented.
Guardrails, schema validation, and human review constrain unsafe or uncertain AI output.
Public-data provenance, source licensing, privacy posture, and limitations stay visible.

Engineering Highlights

Built local-first data platforms with deterministic sample pipelines, analytical models, APIs, dashboards, CI, and guardrails.
Implemented RAG triage workflows with grounded generation, citation checks, offline evaluation, and safe fallback behavior.
Designed ML/MLOps workflows with rolling-origin backtesting, interval monitoring, experiment tracking, and registry-style metadata.
Developed Thai NLP governance workflows with confidence routing, explainability metadata, monitoring, and active-learning queues.
Built privacy-aware multimodal extraction apps with strict schemas, local-first storage, and provider-routing guardrails.

About

Practical engineering across models, APIs, data, interfaces, and deployment.

I am an AI Engineer at Seagate Technology building practical GenAI automation, internal engineering tools, and backend data workflows. My public projects demonstrate production-style thinking across retrieval, multimodal extraction, document analysis, forecasting, and explainable ML.

I use personal projects to show the parts that are often missing from AI demos: input validation, provider routing, safe fallback, measurable evaluation, honest limitations, and interfaces people can actually use.

How I work

1Start with the workflow and failure modes, not the model demo.
2Keep provider boundaries, validation, and fallback behavior explicit.
3Treat deployment, evaluation, and clear limitations as part of the product.

Based in Nakhon Ratchasima, Thailand. Targeting AI Engineer, Data Engineer roles.

Technical Focus

Core engineering areas demonstrated by the public project evidence.

Data Engineering

Python
SQL
DuckDB
dbt-style modeling
Data validation
Reproducibility

AI Systems

RAG
Retrieval
Provider routing
Guardrails
Offline evaluation
Multimodal extraction

ML/MLOps

Backtesting
Interval monitoring
Experiment tracking
Model registry metadata
scikit-learn
PyTorch

Full Stack

FastAPI
React
Next.js
TypeScript
Pydantic
REST APIs

Thai AI/Data

Thai NLP
Buddhist Era normalization
Bilingual evidence UI
Confidence routing
Public data provenance

Experience

AI engineering, full-stack delivery, automation, and data-infrastructure work.

Jan 2026 - Present

AI Engineer

Seagate Technology

Builds practical GenAI automation and internal engineering workflow tooling while supporting backend and data-infrastructure work.

Responsibilities

Develop multi-agent workflows for requirements analysis, debugging, testing, and pull-request preparation.
Build and maintain internal C# engineering software.
Support time-series storage and Kafka-to-TSDB ingestion validation.

Achievements

Applied agentic AI patterns to repeatable engineering workflows.
Contributed to VictoriaMetrics TSDB setup and ingestion verification.
Connected AI application work with operational backend and data concerns.

Technologies

GenAI agents
C#
VictoriaMetrics
Apache Kafka
Apache Flink
Java

Sep 2023 - Jan 2026

Full-stack Developer

WANG CORPORATION CO., LTD.

Delivered full-stack applications and internal automation tools spanning computer vision, data extraction, civic-safety systems, multilingual meeting workflows, and operational web applications.

Responsibilities

Designed React and Next.js interfaces backed by APIs and relational data.
Built AI-assisted extraction, summarization, and computer-vision workflows.
Delivered internal tools for queueing, maintenance requests, and reporting.

Achievements

Built reusable full-stack and automation patterns across multiple internal products.
Converted unstructured inputs into structured datasets and operational dashboards.
Shipped Dockerized services and data-backed applications for remote stakeholders.

Technologies

React
Next.js
FastAPI
PostgreSQL
Directus
Supabase
Docker

How I Can Contribute

Practical delivery areas supported by the public projects and current engineering experience.

From uncertain inputs to a deployable workflow

I can contribute across AI-enabled applications, retrieval workflows, structured extraction, backend APIs, data systems, and the React interfaces that make those systems understandable.

Open to

AI Engineer roles
Full-stack Developer roles
Software Engineer roles
Freelance AI automation projects
MVP builds for AI-enabled products

AI application prototypes

Build working AI product prototypes that connect user input, model calls, structured output, and a usable web interface.

AI Resume MatcherReceipt AI Expense TrackerCustomer Support RAG Triage Agent

RAG and document-grounded chat

Create retrieval workflows for documents, embeddings, vector search, chat history, and streamed assistant responses.

Customer Support RAG Triage AgentThai Procurement Intelligence

Multimodal extraction workflows

Turn images, receipts, screenshots, product photos, and PDFs into normalized records that can be searched or analyzed.

Receipt AI Expense TrackerAI Resume Matcher

Full-stack web applications

Ship recruiter-ready web apps with Next.js, React, TypeScript, API routes, backend services, and deployment setup.

Next.jsReactTypeScriptFastAPIVercelSupabase

Backend APIs and data systems

Design API flows for ingestion, validation, search, analytics, persistence, and AI service integration.

FastAPIPostgreSQLpgvectorSQLAlchemySupabase

Project evidence is linked to public GitHub repositories and deployments. Employment claims are limited to the confirmed resume and profile material used for this portfolio.

Email Me LinkedIn

Let’s Connect

Direct contact, public engineering evidence, and an updated resume.

Interested in practical AI work? Let’s talk.

Email is the fastest way to reach me. GitHub contains the implementation evidence, LinkedIn covers current experience, and the resume provides a concise career summary.

Email me Resume GitHub LinkedIn