AI Engineer LLM & Applied AI

~2 years building AI solutions end-to-end — from model experimentation to integrating AI into backend systems and APIs. Focused on LLM applications, RAG, document extraction and computer vision, with backend experience in Python and Golang.

View Projects Get In Touch

from MyPortfolio import Jirapa
def introduce_myself():
  name = "Jirapa Ramjan"
  role = ["AI Engineer", "Backend Developer"]
  focus = ["LLM", "RAG", "Document AI", "Computer Vision"]
  print(f"Hi, I'm {name} 👋")
  print(f"→ Role: {', '.join(role)}")
  print(f"→ Focus: {', '.join(focus)}")
if __name__ == "__main__":
  introduce_myself()

Output :

Hi, I'm Jirapa Ramjan 👋 
→ Role: AI Engineer, Backend Developer 
→ Focus: LLM, RAG, Document AI, Computer Vision

Technical Expertise

💬

LLM & RAG

RAG chatbots, Hybrid Search (Vector + BM25), Chunking & Embedding, Prompt Engineering, Structured JSON Output

Qwen LangChain Elasticsearch LLM Router

📄

Document AI & OCR

Information extraction from invoices and documents, Vision-Language Models, schema validation before handing data to other systems

PaddleOCR Tesseract Qwen-VL Pydantic

👁️

Computer Vision

Image Classification, Object Detection, Semantic Segmentation (medical imaging: MRI, Chest X-ray)

PyTorch YOLOv8 segmentation_models_pytorch

See notes ↗

🖼️

Image Processing

Image Enhancement (CLAHE), Filtering, Noise Reduction, Data Augmentation

OpenCV scikit-image Pillow

See notes ↗

🗣️

NLP

LLM-based Sentiment Analysis for Thai text (including sarcasm), concurrent batch processing with async requests

Qwen httpx asyncio Pandas

⚙️

Backend & Data

Serving AI models through APIs, RESTful services, ETL pipelines and SQL for BI reporting

Python FastAPI Golang SQL Docker

Experience

AI Engineer · Internet Thailand PCL (INET)

Aug 2024 – Aug 2026

Generative AI & AI Solutions: Collaborated on an enterprise AI Chatbot using Eino and LangChain. Supported the setup of RAG pipelines and used LLM Router APIs to route requests, optimizing response accuracy and resource efficiency.
Document AI & OCR Automation: Applied LLMs (Qwen2VL) with fine-tuning and Prompt Engineering to extract key information for Tax Bill OCR automation, plus a validation and error-handling layer to ensure data accuracy before ERP ingestion.
Computer Vision (Medical & Industrial): Developed and deployed YOLOv8 and Image Segmentation models, including automated wooden log counting for logistics and disease pattern analysis from medical X-ray imagery.
Data Engineering (BI Team): Designed and managed ETL processes and wrote complex SQL based on business logic to transform raw data for BI reporting.
Frontend Development (POC): Built a Vue.js prototype that lets users upload images for AI analysis, with a dashboard visualizing the processed data and key metrics.

LLM RAG Eino LangChain Qwen2VL YOLOv8 SQL / ETL Vue.js

AI Work Highlights

🤖

Enterprise Knowledge AI Agent

Internal RAG chatbot for company documents. Handled data preparation, chunking and embeddings, with Hybrid Search (Vector + Keyword) on Elasticsearch to retrieve Thai text and exact terms like project codes accurately.

RAG LangChain Eino Elasticsearch

🧾

Tax Bill OCR Automation

OCR + Qwen pipeline that turns tax invoices into structured data, with Regex and arithmetic (VAT) validation before ERP ingestion.

> 95% field-level extraction accuracy

Qwen2VL OCR Pydantic FastAPI

📤

Smart Document Upload

API that accepts PDF and images, converts and compresses them, then uses Qwen-VL to extract key fields into validated JSON.

Qwen-VL pdf2image Pillow FastAPI

💭

LLM Sentiment Analysis

Classifies customer feedback as Positive / Negative / Neutral with short reasons, handling Thai sarcasm. Uses async concurrent requests for high throughput.

Qwen httpx asyncio

🩻

Medical Image AI

Disease pattern analysis on medical imagery — brain tumor detection on MRI with YOLOv8 and pneumothorax segmentation on Chest X-ray (U-Net, U-Net++, DeepLabV3+).

PyTorch YOLOv8 U-Net OpenCV

🪵

Wooden Log Counting

Industrial computer vision for logistics — detects and counts wooden logs automatically from images to replace manual counting.

YOLOv8 Object Detection OpenCV

Hackathons & Achievements

2026

AI Engineering Hackathon (Microsoft / AIAT) 🏅 Honorable Mention

"SmartFactory TwinOps AI" on Azure — integrated AI Multi-Agent, Digital Twin and RAG to automate maintenance work orders from real-time sensor monitoring.
2023

Gaskathon (PTT)

Applied ML techniques to predict waste gas volume for sustainable resource management.
2022

Kubota Hackathon

Pitched an AI-driven solution to optimize agricultural land usage.
📄

Mobile Mental Health App (Project & Publication)

Lead full-stack developer (Flutter, PHP, MySQL) for a cross-platform assessment tool. Presented at the 33rd TMI and 13th NCMI Conferences.

Personal Projects

NeuroClassify AI

Upload a brain MRI image and classify it into one of four tumor categories using a deep learning model served via API.

Python PyTorch YOLOv8 React

Live Demo ↗

GitHub

Travel Memory Pinboard

Pin provinces, national parks and countries you've visited on an interactive map, with photos, notes, dates and trip plans.

React TypeScript Tailwind CSS Leaflet

Live Demo ↗

GitHub

Special Gift Website

Personalized gift websites for someone special — Secret Q&A, Memory Timeline and Map Memory templates.

HTML CSS JavaScript

Live Demo ↗ 🔒 Private repo

View All Projects ↗

About me

Personal Information

Name : Jirapa Ramjan

Nickname : Prae

Age : 25 years

Gender : Female

Date of Birth : 3 June 2001

Education

B.Sc. Computer Science, Silpakorn University

2020-2024

GPAX : 3.25

Work Experience

AI Engineer · INET

Aug 2024 - Aug 2026