Peerapon Wechsuwanmanee

• Spearheaded the multi-agent AI system that won the company Hackathon 2025 ('Technology Love' category, among hundreds of developers), auto-resolving common software-development issues; now scaled company-wide, estimated 5%+ productivity gain per developer.
• Architected a multi-agent extension to the store-employee chatbot - natural-language task creation via an MCP tool over a Google ADK agent-to-agent architecture - placing 3rd at Google Agent Factory 2026.
• Owned end-to-end delivery of 4 Generative AI products, from architecture to production, for back-office and in-market use cases - serving thousands of German retail markets and delivering directly to business and operational teams.
• Designed the foundational scalable, observable GenAI/ML architecture (Kubernetes, Prometheus, Grafana, Loki) with zero-downtime deployment and production-grade monitoring across thousands of markets.
• Led design and delivery of a few-shot RAG service (LangChain, Qdrant, FastAPI, Snowflake) that auto-generates personalized coupon copy by aggregating master-data offers across differing grammage and wording (e.g. 'Cola 250-1500 ml'), with a human-in-the-loop correction loop feeding curated examples into later generations - replacing manual weekly copywriting after an in-house migration off a third-party loyalty platform.
• Designed and built a multi-agent GenAI system (FastAPI, Snowflake, pgvector) mapping weekly on-sale items to seasonally-relevant recipes (Easter, Christmas, summer BBQ) with explainable reasoning for marketing materials, eliminating hours of manual item-to-recipe matching per week.
• Led the productionization of a hybrid keyword/semantic, multi-modal internal-document search engine for thousands of markets - owned LogScale-to-Loki logging migration, Prometheus/Grafana metrics, zero-downtime Kubernetes autoscaling, and reliability/Ops work others avoided - replacing a third-party keyword search and cutting employee document lookup from hours to under 5 minutes.
• Architected an agent-to-agent (A2A) orchestrator chatbot (Google ADK, MCP, Kubernetes, LangSmith) routing market-employee queries to specialized agents (semantic document search, task-system reader) as a one-stop in-market assistant.
• Pioneered agentic-engineering adoption beyond own team - presented best practices at the internal Agentic Engineering Community of Practice.

• Pioneered an LLM-based document-summarization MLOps proof-of-concept, building the reusable training/serving accelerator later adopted as an internal delivery template.
• Developed a PyTorch computer-vision defect-detection system for steel-surface production lines, classifying surface anomalies from line imagery.
• Engineered an end-to-end PySpark predictive-maintenance pipeline for the automotive sector - data extraction through feature engineering to a Streamlit decisioning dashboard - directly enabling the client's maintenance-cost reduction.
• Spearheaded extension and optimization of a complex Scala/Spark ETL pipeline for automotive IoT analysis, processing terabytes of sensor data with improved throughput and reliability.
• Delivered a long-stalled ETL project 2x ahead of estimate by implementing the missing domain equations and re-optimizing the data-processing workflow.

• Led Speech-to-Text product team to build speaker diarization, speaker recognition, and speech recognition models.
• Built end-to-end ASR pipeline (data acquisition → fine-tuning → deployment) reaching 8% Character Error Rate on the in-domain Thai test set versus 11% for Google Speech Recognition - outperforming the incumbent on a low-resource language in target domains.

• Conducted and coordinated international research projects in computational mechanics and damage modeling of advanced high-strength steels.
• Published 10 peer-reviewed papers (125+ citations); first-authored 'Influence of surface roughness on cold formability in bending processes' in International Journal of Material Forming (2021).
• Developed multiscale numerical models combining finite element simulation with experimental material characterization.
• Collaborated across institutions including Aalto University (Finland) and TU Delft (Netherlands) on EU-funded research.

Semantic Product Search Engine

• Researched and evaluated search engine options for a second-hand e-commerce platform, selecting Meilisearch for its hybrid semantic/lexical search, bilingual Thai/English support, and cost profile at scale.
• Designed and deployed a real-time PostgreSQL → Meilisearch sync pipeline using logical replication and MeiliBridge, enabling sub-second product index updates without changes to the host application.
• Tuned search relevance through ranking rules, typo tolerance, embeddings, and query strategies (type-ahead, faceted filtering, brand relaxation), improving product discoverability.
• Contributed compatibility patches to the open-source MeiliBridge Rust codebase to unblock production adoption.

Motorbike Condition Evaluation

• Built a multi-modal ML regression pipeline fusing 39-dimensional MFCC audio features (extracted from inspection videos via librosa), Thai-language BERT embeddings (WangchanBERTa), and categorical inspection inputs to predict a 0-100 motorcycle condition score.
• Benchmarked 5 regression algorithms (ElasticNet, Random Forest, Gradient Boosting, SVR, XGBoost) with scikit-learn RandomizedSearchCV hyperparameter tuning; serialized the best-performing pipeline for cloud-hosted inference.
• Engineered an audio feature extraction pipeline that converts MP4 inspection videos to log-mel spectrograms, computes MFCC + delta + delta-delta coefficients, and normalizes across axes to produce a stable 39-feature vector.
• Deployed a production FastAPI + Streamlit application with object-storage video storage, PostgreSQL result logging, and Docker — submitting a video returns a scored risk assessment in real time.

Credit Risk Assessment

• Performed exploratory analysis and built a reproducible data-cleaning and transformation pipeline on a raw credit dataset.
• Ran model selection across candidate classifiers with per-segment personalized feature selection to sharpen credit-default discrimination.
• Delivered an interactive Streamlit decisioning UI to score applicants and inspect the drivers behind each prediction.

Credit Default Prediction

• Built a full-lifecycle credit-default MLOps system: EDA, automated data-cleaning pipeline, MLflow experiment tracking, blue-green model selection, and Prefect-orchestrated automatic retraining.
• Ran a profit-optimized threshold simulation achieving a 30% profit increase on out-of-sample data versus a baseline cut-off policy.
• Deployed a production-grade, scalable FastAPI inference service serving real-time default predictions, with retraining triggered automatically as new repayment data arrived.

Customer Feedback Classification

• Reduced case resolution time by 18% by building an automated Thai-language text classifier that routes incoming feedback to the correct service team.
• Implemented a RAG pipeline using OpenAI embeddings and ChromaDB to retrieve similar labeled examples as few-shot context, improving classification accuracy without retraining.
• Packaged the system as a Dockerized FastAPI service, enabling new training examples to be added at runtime via CSV upload without redeployment.

Built a multi-agent extension to REWE's centralized store-employee chatbot enabling natural-language task creation via a FastMCP MCP tool, on a decoupled Google ADK agent-to-agent (A2A) architecture.

Developed multi-agent AI system for automatic software development issue resolution.

Built real-time anomaly detection deep learning model for predictive maintenance of electrical drives.

Built ML algorithms to investigate microstructure, fracture toughness, and early-stage fatigue of steels in nuclear power plant from non-destructive evaluation (NDE) measurements. Collaborated with international teammates in agile environment.

Built computer vision model for 42-class e-commerce product classification and NLP model for product rating prediction in English from Malaysian/Indonesian websites.

Improved predictive maintenance accuracy from turbocharger inspection data.

Built ML model to predict success rate of part extraction from laser machine given geometry features and machine parameters.

Competed in the international RoboCup Small Size Soccer League representing Kasetsart University.

Influence of surface roughness on cold formability in bending processes: a multiscale modelling approach with the hybrid damage mechanics model

International Journal of Material Forming, 14(6) (2021) - 11 citations DOI

Numerical Evaluation of Surface Roughness Influences on Cold Formability of DP Steel

13th Intl. Conf. Numerical Methods in Industrial Forming Processes (NUMIFORM) (2019)

Modeling the Surface Roughness Influence on the Hole Expansion Ratio of Multiphase Steel

38th Intl. Deep Drawing Research Group (IDDRG) Annual Conference (2019)

Scalebridging Approaches to Assess the Edge Crack Sensitivity of Multiphase Steels

13th Intl. Conf. Numerical Methods in Industrial Forming Processes (NUMIFORM) (2019)

A Numerical Approach to Evaluate Roughness Effects on Localization and Damage in Sheet Materials

37th Intl. Deep Drawing Research Group (IDDRG) Annual Conference (2018)

Finite Element Model in Abrasion Analysis for Single-Asperity Scratch Test

Fracture Fatigue and Wear, Springer (2018)

Multi-objective Optimization of Multi-scale Finite Element Analysis for Wear Resistant Steel

U.S. National Congress on Computational Mechanics (2017)

Prediction of Plasticity and Damage Initiation Behaviour of C45E+N Steel by Micromechanical Modelling

Materials and Design (2017)

The Modelling Scheme to Evaluate the Influence of Microstructure Features on Microcrack Formation of DP-steel

Computational Materials Science (2014)

Development of the Modified Kinematics for a Wheeled Mobile Robot

Intl. Technical Conf. Circuits/Systems, Computers and Communications (2010)

Kinematics Simulation by using MSC.ADAMS for a 6-DOF Industrial Robot Arm

Thailand Simulation and Modeling Conference (2010)

• Designed and taught a weekend cohort of ~40 IT and non-IT professionals end-to-end AI product development, from model to deployed service.
• Topics: Python, Machine Learning, FastAPI, LLM API, Streamlit, Docker, Linux, AWS.

• Led a weekend cohort of ~100 data enthusiasts into data-engineering careers through lectures and hands-on workshops.
• Topics: Basic Data Engineering, Python, SQL, AWS, Git, Workflow Orchestration.
• Prepared and managed AWS cloud infrastructure for data engineering, data science, and data analytics sections.

• Co-organized meetup with Seldon at Data Reply offices in Munich for data enthusiasts.
• Presented talk on 'Big Pictures and Use-cases in Predictive Maintenance'.
• Published recap article on Medium (DataReply publication).

• Provided free digital education for locals, newcomers, and refugees in Munich.
• Supervised and guided students for their final projects.

• Established the Thai language team and led localization of 2 out of 9 chapters of the HuggingFace NLP course.

• Provided hands-on lectures to 450 online students - Intro to Python, NumPy/Pandas, Feature Engineering, Model Deployment.
• Supervised students' end-to-end Computer Vision, NLP, and Time-Series Analysis projects.

• Provided weekly supervision and consultancy on data science projects for career switchers.
• Donated all compensation to charitable institutions (UNICEF, Wikimedia, ProVeg).