GitHub profile  ·  LinkedIn profile  ·  Email Kun Ming  ·  Browse all projects

AI systems · Data products · Full-stack delivery · Responsible engineering

About  ·  Approach  ·  Skills  ·  Experience  ·  Projects  ·  Learning  ·  Contact

Designed to be read, not merely viewed. Every essential detail is available as native text; the artwork is supplementary. Motion stops when reduced motion is requested, and the expandable sections use keyboard-operable GitHub controls. Read the accessibility notes →

I’m Goh Kun Ming, a Year 3 student in the Diploma in Applied AI & Analytics at Singapore Polytechnic, based in Singapore. I am developing toward the kind of AI practitioner who can move between research questions, data, models, software systems, and the people who eventually have to understand or use what was built.

My technical experience spans data engineering and analysis, machine learning, deep learning, model evaluation, computer vision, conversational and retrieval-based AI, full-stack development, and cloud delivery. I enjoy working across those boundaries because the most interesting problems rarely stop at a notebook: they continue into data contracts, APIs, interfaces, testing, deployment, monitoring, documentation, and governance.

My research interests include intelligent systems, AI safety and governance, quantum computing, and complexity science. Across those areas, I keep returning to the same questions: Is the problem framed well? Is the evidence trustworthy? Can the system explain where its answer came from? Does it remain useful outside the happy path? Can someone else inspect, reproduce, and improve it?

Longer term, I hope to contribute to AI academia while continuing to build applied systems. Teaching and mentoring matter to that direction: I want to become good not only at developing technical ideas, but also at making them understandable to other people.

• Based in: Singapore
• Studying: Year 3, Diploma in Applied AI & Analytics at Singapore Polytechnic
• Technical range: data pipelines → analytics → ML/DL → RAG → APIs → interfaces → testing and delivery
• Research interests: intelligent systems · responsible AI · quantum computing · complexity science
• Working style: evidence-led, systems-minded, documentation-conscious, and curious about first principles
• Long-term direction: AI research, academia, teaching, and mentoring

The short version: I do not want the model to be the whole story. I want the question, data, decisions, interface, tests, limitations, and handover to be visible too.

1. Start with the decision, not the algorithm. I first ask who needs the result, what they need to decide, and what a useful answer would look like.
2. Treat data quality as part of the model. Provenance, missingness, leakage, definitions, and validation matter before architecture selection.
3. Build a baseline before adding complexity. A sophisticated system should earn its additional moving parts through evidence.
4. Connect the whole path. I think about how ingestion, modelling, retrieval, APIs, interfaces, roles, and deployment affect one another.
5. Design for failure as well as success. Robustness, uncertainty, access control, fallbacks, and limitations deserve explicit treatment.
6. Leave an inspectable trail. Tests, documentation, evaluation artifacts, and handover notes make the work easier to trust and extend.

• A research-and-systems perspective. I like understanding why an approach should work, then following it through the architecture needed to make it useful.
• Questions before assumptions. I try to make the target, constraints, evaluation criteria, and failure modes explicit before optimising a solution.
• Visible trade-offs. I would rather document a limitation or uncertain result than hide it behind polished presentation.
• Attention to the less glamorous layers. Data cleaning, access rules, tests, deployment notes, and maintenance documentation are part of the build.
• Communication as an engineering skill. I use diagrams, model cards, data contracts, examples, and structured documentation to make technical decisions easier to review.
• Curiosity beyond one speciality. My work moves across algorithms, analytics, ML, deep learning, language systems, vision, web products, cloud services, and emerging computation.

My recent experience has given me practice at three different scales: an institutional AI product, a multi-role data and workflow platform, and an early-stage recommendation concept. Together they strengthened my ability to move between research, engineering, product questions, and verification.

These are concise highlights from my public LinkedIn experience. The numbers describe documented build scope and verification—not business-impact claims.

The projects below are public, inspectable, and chosen to show range: research experiments, data products, and end-to-end AI applications.

How can model experimentation stay reproducible from configuration to evaluation?

How does a large dataset become a decision someone can understand?

What turns an AI prototype into a system another person can actually use?

My recent learning has been deliberately broader than a single model family. I am building a foundation for studying AI as both a technical system and a human responsibility.

• Responsible AI and governance: fairness, accountability, transparency, privacy, safety, data governance, and responsible decision-making.
• Deep learning: neural-network fundamentals, optimisation, evaluation, multi-input and multi-output architectures, embeddings, and practical model development.
• Quantum computing: qubits, superposition, entanglement, quantum gates and circuits, Qiskit, and the relationship between classical and quantum approaches.
• Complex systems: emergence, resilience, networks, tipping points, agent-based modelling, and how interconnected systems behave under change.
• Research communication: explaining assumptions, methods, evidence, limitations, and implications in a form that technical and non-technical readers can follow.

Selected public credentials:

• Quantum Computing For Everyone — An Introduction, Fractal, Jul 2026
• Ethical AI — AI Essentials For Everyone, University of Cambridge, Jun 2026
• Introduction to Complexity Science, Nanyang Technological University Singapore, Jun 2026
• Advanced Deep Learning with Keras, DataCamp, Mar 2026
• Fundamentals of Deep Learning, NVIDIA, Feb 2026

View the public credential record →

This is not a mock-up. The shape is generated from a parametric (2, 3) torus-knot tube: 1,728 vertices and 3,456 triangular faces. It represents how I prefer to work—question, data, baseline, experiment, product, and sharing are distinct strands, but the useful artifact is one continuous system.

Open the interactive 3D viewer  ·  Download the OBJ model  ·  Read the generator

I like ambitious ideas, but I trust them more when the path is explicit:

Question → data → baseline → experiment → product → evidence → handover

• How can retrieval and conversational AI remain grounded, useful, and honest about uncertainty?
• How should evaluation change when an AI system includes data pipelines, retrieval, interfaces, and human decisions—not only a model?
• How can multimodal systems combine structured data, language, images, and sensor-like signals without losing interpretability?
• Where can quantum or hybrid methods be compared with classical baselines in a bounded, reproducible way?
• How can technical tools also become better learning objects for students, practitioners, and decision-makers?

I am most interested in work where the problem is not fully solved by choosing a model: the data needs structure, the system crosses multiple layers, reliability matters, and the final result must be communicated clearly. That includes applied AI, data products, research engineering, evaluation, responsible-AI work, and technically grounded experimentation.

If you are evaluating my work for a role, exploring a collaboration, or simply working on careful AI systems, I would be glad to talk.

Email me  ·  Connect on LinkedIn  ·  Browse every repository

_{Curious by default. Clear by design.}