Flora Liu Hardware Engineer

05 About

About

I am Flora Liu, a hardware engineer at Amazon, working on the product integrity side of new product introduction.

Mechanical engineering gave me a physical foundation. Reliability work, product development and research taught me to connect that foundation to data, and then to decisions. That turned out to be the harder half.

I have been useful in three different seats: a third-party laboratory that produces evidence, a system company that builds and pushes back, and a global brand that has to decide and live with the consequence. I write because the interesting part of engineering is what happens in between, and almost nobody writes that down.

Lately my curiosity goes into two things: making reliability evidence legible to the people who carry the business risk, and using agents to remove coordination friction in hardware development without handing over the judgement.

Portrait of Flora Liu. Download résumé (PDF)
5.1

Experience

Amazon

Hardware Engineer · Product Integrity

Mar 2025–present

Joined as Hardware Development Engineer Intern; engineer since Mar 2026.

  • Hardware qualification across the full NPI lifecycle: EVT, DVT, PVT, HVT through mass production.
  • Reliability statistics for risk-based quality decisions: lifetime modelling, distribution fitting, and uncertainty quantification rather than point estimates.
  • Built a test-data platform (MySQL + dashboards) so reliability evidence could be queried instead of reassembled by hand.
  • AI-assisted failure analysis on test and field-return data, in collaboration with manufacturing partners.
  • Automated the coordination layer of NPI (issue tracking, status queries, reporting) so engineering time goes to judgement, not status chasing.

TSMC

Campus Ambassador

Mar 2025–Mar 2026

  • Represented TSMC to students at National Taiwan University: campus job fairs and recruiting sessions for internships, pre-hire offers and R&D alternative service.

Delta Electronics

R&D Engineer Intern · Lab for Digital Twin-Based Optimization

Jul–Aug 2024

  • Built modular Python tooling to automate parametric modelling and FEA execution, shortening the design-to-analysis loop.
  • Wrapped it in an interface non-programmers could use, then iterated with the engineers who actually ran it.

SGS Taiwan

Reliability Intern · Reliability Laboratory

Sep 2023–Apr 2024

  • Ran vibration, shock, thermal, ingress and HALT testing; instrumented rigs with LabVIEW and DAQ.
  • Cross-checked physical measurements against FEA to test whether the two stories agreed, and reported honestly when they did not.
  • Introduced an FEA-supported workflow so clients received engineering insight, not just a pass/fail certificate.
5.2

Education

National Taiwan University

M.S., Engineering Science and Ocean Engineering (Electrical & Electronic)

Sep 2024–Jul 2026

First-author paper, presented orally at SPIE Smart Structures + NDE 2026, Vancouver: load-carrying and stability improvement of a miniature ultrasonic piezoelectric plate motor.

Proc. SPIE 13949 · doi:10.1117/12.3090434 ↗

National Cheng Kung University

B.S., Mechanical Engineering

Sep 2020–Jun 2024

Research: sliding-mode control system for vibration control.

5.3

Skills

Named specifically, because "statistical analysis" is a claim and these are checkable.

Reliability & statistics
Weibull and lognormal life fitting B10 life, MTTF Bootstrap and Monte Carlo uncertainty bounds KS goodness-of-fit Accelerated ageing Annual return rate (ARR)
Test & instrumentation
Vibration, shock, thermal, ingress and HALT testing LabVIEW and NI DAQ SCPI instrument control Camera-based motion measurement Simulink, LTspice, Keil
Simulation & CAD
Ansys and COMSOL FEA, scripted Simulation-to-test correlation SolidWorks, AutoCAD
Data & software
Python (pandas, NumPy, PyTorch) MATLAB C++ SQL and MySQL Docker Git
AI workflow
LLM agents MCP LangChain, LangGraph AWS
Languages
Mandarin (native) English (professional, TOEIC 770)
5.4

Certifications

5.5

Selected work

Public repositories. Work done at Amazon is not represented here and will not be. What I can share about it lives in the written pages instead.

Closed-loop control of a miniature physical system

Research instrumentation

A measurement and control rig built end to end for my thesis work: multithreaded camera tracking at 120 fps, synchronised multi-channel waveform excitation over SCPI, and closed-loop attitude correction.

  • Separate acquisition, processing and display threads behind a lock-protected shared state
  • Kalman and exponential-moving-average filtering; thin-plate-spline lens distortion correction
  • PID attitude correction with differential-voltage steering across two channels
  • Automatic scale calibration, CSV export, and generated trajectory/velocity figures

Repository →

Test-to-failure analysis toolkit

Reliability statistics

Fits life distributions to test-to-failure data and generates a report: survival-function comparison, cumulative failure curves, and distribution contribution. Anonymised datasets only.

Repository →

3 more projects: data pipelines, machine learning, undergraduate archive

Neural-network surrogates for flow regression

Machine learning

Regression of Couette and Hagen–Poiseuille flow fields, with a deliberate architecture ablation (baseline against deeper, wider and different-activation variants) to see what mattered.

Repository →

Instrument data pipeline

Data engineering

Automated ingest of logger spreadsheets into MySQL with hash-based duplicate protection, scheduled backup and cleanup, rolling statistics and alerting: the unglamorous layer that makes measurement data usable.

Repository →

Undergraduate portfolio

Archive

Mechanical engineering coursework and projects from NCKU: kinematics, mechanism design, mechanical drawing, numerical analysis, robot design, instrumentation and reliability.

Repository →

5.6

Outside the lab

I swam competitively, and it shaped how I work more than any course did: measure honestly, adjust continuously, and stay consistent long enough for it to compound. Reasonable training for reliability engineering, where nothing interesting happens quickly.