P.P.
Bangkok --:--
CS @ Thammasat University Class of 2027

Pitawat Piruksa

I build agentic retrieval systems that let language models reason over buildings, graphs and messy engineering data, and I measure what they cost to run.

01About

I'm a fourth-year Computer Science student at Thammasat University, Thailand, specialising in AI engineering, agentic systems and cloud infrastructure.

Portrait of Pitawat Piruksa

I build end-to-end hybrid RAG pipelines that combine vector databases with knowledge graphs, implement context caching and LLM query routing, and deploy serverless backends on AWS. I'm looking to build production AI systems using my machine learning and software engineering background.

Studying
B.S. Computer Science, Thammasat University
GPA
3.43 / 4.00
Period
Aug 2023 — Aug 2027 (expected)
Looking for
Building production AI systems
02Experience
Jul — Aug 2026

Academic Research Intern

Kansai University, Suita Mirai Campus · Osaka, Japan

Optimised retrieval-augmented generation (RAG) architectures with Prof. Kohei Ichikawa. Designed experimental frameworks to evaluate trade-offs across retrieval depth (vector top-k), explicit context caching and LLM query routing.

result 95.7% fewer uncached input tokens through explicit context caching; first-author paper accepted at IEEE ICA 2026.

  • Python
  • RAG
  • Explicit context caching
  • LLM query routing
  • Vector retrieval
03Projects
04Publications
  1. [1]

    Retrieval Depth, Token Usage, and Query Routing in Agentic Information Retrieval from Building Descriptions

    P. Piruksa, P. Rattanatamrong, K. Ichikawa, J. A. B. Fortes, V. Likhitruangsilp

    IEEE ICA 2026 · International Conference on Agents · accepted

  2. [2]

    Agentic Hybrid RAG for Spatial Reasoning in BIM Using Knowledge Graphs and Vector Retrieval

    P. Piruksa, W. Putthividhya, P. Rattanatamrong

    ECTI-CARD 2026 · Nakhon Ratchasima, Thailand · proceedings

05Toolbox
AI & language models
  • Agentic workflows
  • RAG
  • LangChain
  • Neo4j
  • FAISS
  • Context caching
  • Prompt engineering
  • Gemini
  • PyTorch
  • TensorFlow
  • scikit-learn
Engineering & cloud
  • AWS
  • AWS Lambda
  • AWS S3
  • Docker
  • Flask
  • REST APIs
  • Git
  • GitHub
Languages
  • Python
  • C
  • C++
  • Java
  • SQL
  • Cypher
  • SPARQL

06 — Get in touch

q.pitawat@gmail.com

Open to research collaborations, lab visits and conversations about agents that retrieve things.