Skip to my experience
Huang LinMachine Learning Engineer & Researcher

Huang Lin — Machine Learning Engineer & Researcher

HUANG / EXPERIENCE MAP

Auto journey · Pause to explore.

The journey, in words

Education, experience & current skills

Back to the cities ↑

01 / Guangzhou · The foundations

Engineering Foundations

B.Sc. in Electrical Engineering · South China Agricultural University · 2017

Foundations in MATLAB, mathematics, statistics, and probability developed during my undergraduate studies.

2017

B.Sc. in Electrical Engineering

South China Agricultural University. Awarded a Second Prize Scholarship.

Full experience ↗Education & CV ↗

Professional experience · China Telecom

SQL reporting & operations analytics.

Sales Operations Analyst · July 2017 – January 2021

Built SQL dashboards for KPIs and sales trends, maintained data quality, and turned operational questions into repeatable reporting workflows.

Full experience ↗

Current skills · Across my experience

From data to a working system.

Software tools, data analysis, model development, and research methods.

Languages
Python · SQL · MATLAB
Modeling
PyTorch · TensorFlow · scikit-learn
Engineering
Data pipelines · Feature engineering · Model training · Evaluation · Optimization

What I explore

Applied machine learning, AI applications, algorithm design, structured deep learning, AI systems evaluation, and representation learning.

Explore all projects ↗Certificates ↗

02 / Regina · Doctoral research

Learning from signal data.

Ph.D. in Engineering · University of Regina
January 2021 – May 2026
Research in machine learning for signal-based indoor localization.

Research problem: estimate indoor position from high-dimensional WiFi fingerprints while controlling matching cost and sensitivity to signal variation.

Methods: similarity-based candidate selection, signal representation learning, structured attention, and knowledge distillation. The papers and repositories below document the methods and experiments.

Group Matching Method

Reduce the search space for RSS fingerprint matching.

Paper ↗Code ↗

LSTM–MLP knowledge distillation

Connect temporal modeling with lightweight inference.

Paper ↗Code ↗

Saskatchewan Innovation and Excellence Graduate Scholarship, 2022 and 2025.

All publications ↗Localization projects ↗

Montreal · Ericsson

ML pipelines for signal data.

Machine Learning Engineer Intern · Ericsson
March 2021 – April 2023

This Montreal-based role overlapped with my doctoral studies at the University of Regina.

Engineering focus: preprocessing, feature engineering, and similarity-based matching for high-dimensional WiFi data, followed by model training and benchmarking.

  • Python, NumPy, and Pandas workflows for large-scale WiFi datasets.
  • ML and deep learning models using PyTorch, TensorFlow, and scikit-learn.
  • Evaluation and visualization tools for model analysis and technical reporting.
Read the experience ↗5G simulation work ↗

Professional experience · Mercor

Evaluating LLM workflows.

Machine Learning Engineer (Contract) · Mercor
November 2025 – May 2026

Shown with recent work; the Bay Area chapter indicates my current location, not the location of this contract.

Engineering focus: Python pipelines for LLM-driven task execution and reproducibility, evaluation frameworks for agent systems, benchmarking, and analysis of workflow failures.

Questions behind the work

  • Does the output follow the instructions?
  • Can we reproduce the result?
  • Where does the workflow fail, and how do we measure it?
LLM & agent evaluation ↗Engineering tools ↗

Ericsson · Montreal · March 2021 – April 2023

Data & Feature Pipelines

Prepared high-dimensional WiFi datasets with Python, NumPy, and Pandas. Built preprocessing and feature engineering workflows for model training and benchmarking.

Experience ↗All projects ↗

University of Regina · PhD research · January 2021 – May 2026

Efficient Algorithms

Group-based matching reduces the candidate search space for RSS fingerprint localization. The research examines localization accuracy alongside computational efficiency.

Paper ↗Code ↗All projects ↗

University of Regina · PhD research

Representation Learning

Fingerprint transformation, similarity filtering, and adaptive reference selection address the structure of high-dimensional signal data. Structured attention is another research direction documented in the project overview.

Paper ↗Code ↗All projects ↗

University of Regina · PhD research

Knowledge Distillation

The dual-mode LSTM–MLP framework connects temporal signal modeling with lightweight localization inference through knowledge distillation.

Paper ↗Code ↗All projects ↗

Ericsson & University of Regina · Overlapping industry and doctoral work

Model Benchmarking

Developed and compared ML and deep learning models, with evaluation and visualization tools for performance analysis. Investigated signal noise and missing data to understand model robustness.

Experience ↗All projects ↗

Personal project · JavaScript Chrome extension

Software Tooling

Built an application tracker with page metadata extraction, platform detection, notes, and a tracking dashboard. This project demonstrates browser tooling and workflow organization.

Code ↗All projects ↗