Portrait of Syahmi Shamsul
Syahmi Shamsul
Software Engineering & Data Analytics Intern at Cashku · Kuala Lumpur

Machine learning that makes it to production.

Final-year Computer Science student at Taylor's University, specialising in Data Science and AI. I build models, the data work behind them, and the apps that put them in front of people.

👋 About

I'm a final-year Computer Science student specialising in Data Science with an extension in Artificial Intelligence. I care about the whole path from raw data to something people can use: cleaning and auditing data, training and evaluating models, then deploying them behind real interfaces.

Right now I'm a Software Engineering & Data Analytics Intern at Cashku, a Malaysian fintech, where I work on data quality across the platform's fund universe. Before that I led a five-person team building MediVision, a patient flow platform for emergency departments with an XGBoost triage model in production.

I'm looking for a full-time or graduate role in data science, AI, machine learning or software engineering.

📌 At a glance
Status
Open to full-time roles
Roles
Data science · AI · Machine learning · Software engineering
Current
Software Engineering & Data Analytics Intern, Cashku
Education
Bachelor of Computer Science (Honours), Taylor's University
Graduating Feb 2027
Location
Kuala Lumpur, Malaysia
Languages
English, Malay

💼 Experience

Jul 2026 - PresentCurrent

Software Engineering & Data Analytics Intern

Cashku ↗ · Fintech investment platform · Kuala Lumpur

  • Own data-quality investigations across the platform's fund universe: audit coverage with SQL, trace anomalies to root cause, and turn one-off checks into recurring monitoring
  • Cross-reference internal records against external sources to separate pipeline gaps from genuinely missing data, with findings feeding directly into production fixes
  • Work in a Jira-driven agile team alongside product and engineering: run UAT test cases, write precise tickets, and keep meeting minutes and stakeholder updates
  • Report findings daily to the CTO as plain-language, quantified summaries prioritised by business impact
SQLPythonData qualityRoot-cause analysisJiraUATStakeholder comms
Jan 2026 - Apr 2026

Student Assistant, FitFuel Station research project

Taylor's University · INNOVATE Grant

  • Coordinated "FitFuel Station", an AI-driven vending device that customises foodstuff to the user
  • Translated software logic, handled project data and validated the nutrition algorithms
  • Kept body-composition inputs, decision engines and AI-controlled dispensing consistent end to end
PythonData handlingAlgorithm validationResearch coordination

🚀 Projects

Six projects, four with live demos. Open any card for the full write-up, screenshots and an embedded demo.

MediVision AI triage with nurse review
Final Year Project · Full-stack + ML

MediVision: patient flow platform for emergency departments

XGBoost triage classifier served as a microservice, real-time patient tracking and analytics dashboards. Led a team of five and owned the backend and ML integration.

ReactNode.jsMongoDBFastAPIXGBoost
Grad-CAM heatmaps on X-rays
Computer vision · Explainable AI

Bone fracture detection from X-rays

EfficientNetB0 fine-tuned on 4,083 hospital X-rays with CLAHE preprocessing. Grad-CAM shows where the model looks, which also exposed a shortcut-learning issue.

86% accuracy · 0.89 AUC · FracAtlas

TensorFlowOpenCVGrad-CAMGradio
AI-Governance RAG live demo
LLM · Retrieval-augmented generation

RAG over AI-governance law

Question answering over roughly 650 AI laws and policies. Every claim is cited to its source and the system declines when the corpus can't answer. Built without a RAG framework.

5,203 chunks · FAISS · MiniLM · Gemini

FAISSEmbeddingsGeminiPython
Top 15 most important features
Machine learning · Credit risk

Loan repayment risk prediction

Engineered a repayment label and payment-behaviour features from three merged lending datasets, then compared Logistic Regression against Random Forest.

0.814 AUC · 80% bad-loan recall

pandasscikit-learnPlotly
kadku digital business card
Web · Weekend build

kadku: digital business card

Mobile-first card that puts contact details, LinkedIn and resume behind one QR scan. Bilingual in Malay and English, deployed to a custom domain on every push.

HTMLTailwindGitHub Pages
Smart plant monitoring prototype with ESP8266, sensors, relay, pump and LCD
IoT · Embedded

Smart monitoring for houseplants

ESP32 and Raspberry Pi prototype that waters plants automatically and tracks moisture, temperature, humidity and soil pH in real time.

ESP32Raspberry PiPython

🛠️ Skills

Languages and data7

PythonSQLRJavapandasNumPyscikit-learn

Machine learning8

Supervised learningUnsupervised learningClassificationRegressionClusteringRandom ForestXGBoostModel evaluation

Deep learning and LLMs7

TensorFlowKerasPyTorchCNNLSTMTransfer learningRAG

Engineering and tools10

ReactNode.jsFastAPIMongoDBGitJiraGradioHugging FaceTableauColab

✉️ Let's talk

Looking for full-time and graduate roles in data science, AI, machine learning or software engineering. Send a message here and it lands straight in my inbox.

Please enter your first name.
Please enter your last name.
Please enter a valid email address.
Digits, spaces and dashes only.
Please write at least a short message.
I reply to the email address you enter.