HOÀNGAI / ML RESEARCHER
HOÀNG / COMPUTER SCIENCE × AIHCMC · VIETNAM

I build intelligent systems, study how they work, and turn experiments into knowledge.

A Computer Science student working across biomedical language, retrieval, video understanding, and the mathematical foundations beneath machine learning.

Explore my work Read my notes GitHub CV
FIG. 00 — REPRESENTATION STUDY

Signals enter. Structure emerges. A decision becomes inspectable.

Currently studying

Assertion-aware clinical entity linkingUpdated 28.07.26
01

Selected work

Systems are stories about decisions.

Three active records—each showing the question, architecture, constraints, evidence, and unfinished edge.

P–01 / BIOMEDICAL NLPActive research

Clinical Entity Recognition and Linking Pipeline

A Vietnamese clinical NLP pipeline that separates span detection, assertion handling, candidate retrieval, and ontology linking so each source of error can be inspected independently.

  • XLM-R
  • Linear-chain CRF
  • PyTorch
  • BM25
  • Dense retrieval
  • Cross-encoder reranking
Explore the case study
P–02 / VIDEO UNDERSTANDINGExperimental

POV Video Segmentation and Keyframe Selection

A shot-aware preprocessing pipeline for long first-person video: locate visual boundaries, sample representative frames, embed them, and remove near-duplicates without erasing meaningful transitions.

Inspect the pipeline
P–03 / LEARNING LABOngoing

AI Algorithms Learning Laboratory

A structured learning program that moves from mathematical derivation to implementation, tiny verification cases, and visual explanation while preparing for the 2026 Vietnamese Student AI Olympiad.

Open the learning log
02

Research areas

Questions I keep returning to.

A map of active interests, not a claim of mastery. The connective tissue is representation, inference, and evaluation.

01

Biomedical NLP

Turning noisy clinical language into structured, inspectable knowledge.

02

Information Retrieval

Combining lexical recall, learned representations, and careful reranking.

03

Representation Learning

Studying the geometry that makes retrieval and prediction useful.

04

Video Understanding

Finding temporal structure and representative moments in long POV footage.

05

Mathematical Foundations

Deriving the probability, optimization, and linear algebra beneath models.

06

Algorithms

Implementing small reference systems until every state transition is visible.

03

Capabilities

A working vocabulary, not a scorecard.

Tools and methods grouped by how closely they sit to the current work. No invented proficiency percentages.

01

Daily foundations

Core

  • Python
  • PyTorch
  • Machine learning
  • NLP
  • Mathematical reasoning
02

Current research

Active

  • Biomedical NLP
  • Entity linking
  • Information retrieval
  • Structured prediction
  • Error analysis
03

In the build loop

Working

  • Video understanding
  • TypeScript
  • NumPy
  • FFmpeg
  • Temporal sampling
04

Questions in motion

Exploring

  • Representation learning
  • Multimodal models
  • Attention
  • Hybrid retrieval
  • Evaluation design
04

The lab

Inspect the decision, not only the output.

Small interactive instruments for structured prediction, retrieval, and video understanding—built to expose the internal state rather than hide it behind a polished demo.

Enter the lab
05

Technical writing

Ideas become clearer when they survive an explanation.

01

Understanding Viterbi Decoding in Linear-Chain CRFs

How does a global search over exponentially many label sequences collapse into a quadratic-time recurrence without changing the answer?

Mathematical NLP9 min read
02

Practical Error Analysis for Biomedical Named Entity Recognition

When an NER score moves, which boundary, type, context, and annotation failures actually changed—and what should be fixed next?

Biomedical NLP6 min read
03

How TransNetV2 Detects Shot Boundaries

What evidence distinguishes a camera cut from ordinary motion, and how should per-frame boundary probabilities become usable shots?

Video Understanding6 min read
04

Dense Retrieval Versus Lexical Retrieval for Medical Entity Linking

Should a medical entity linker trust shared words, learned semantic proximity, or a deliberately hybrid candidate set?

Information Retrieval6 min read
Browse all notes
06

About

Learning by making the hidden steps visible.

I'm Hoàng, a Computer Science student focused on artificial intelligence, natural language processing, machine learning, and the mathematics underneath them.

This portfolio is also a working record: systems I am building, explanations I needed in order to understand them, and experiments whose conclusions are still allowed to change.

More about how I work
01question02model03test04revision
07

Now / current work

Three questions under active pressure.

A snapshot of what is closest to the keyboard, kept small enough to be honest and current enough to be useful.

01

Biomedical NLP

↗
Latest milestone
Separated recognition, assertion, retrieval, and reranking into inspectable stages.
Next question
Where does linking recall fail before the reranker sees the correct concept?
02

Video Understanding

↗
Latest milestone
Defined a timestamp-preserving path from shot boundaries to keyframe deduplication.
Next question
Can local filtering preserve brief actions near uncertain boundaries?
03

Mathematical Foundations

↗
Latest milestone
Established a derive–implement–interrogate–verify loop for algorithm modules.
Next question
Which examples best expose the difference between max-product and sum-product inference?
Read the full update
08

Contact

Have a difficult problem or an interesting question?

I'm interested in thoughtful conversations about AI systems, research, engineering, and the work of making technical ideas clear.

minhhoang2712006@gmail.com