Technical documentation South San Jose, California

Tao Li.

Senior Technical Writer · Amazon — Annapurna Labs · AWS Neuron SDK

Documentation is detective work: I investigate complex systems, ask engineers the questions customers will ask, and turn the evidence into docs that developers — and AI agents — can trust.

Exhibit AThe method

Twenty-plus years of cases, one procedure.

  1. Investigate

    Read the code, run the product, follow the data. The system itself is the primary source — everything else is hearsay.

  2. Delegate Amended 2025

    The newest procedure in the file: hand the legwork to AI agents — sweeping codebases, tracing call paths, drafting the first pass. Two decades of casework taught me what to look for; the agents widen the search, and I cross-examine every claim before it enters the file.

  3. Interrogate

    Ask engineers the questions customers will ask, before customers have to. The gaps between answers are where the real docs live.

  4. Verify

    Spot-check what I can — run the examples, walk the quickstarts — and route the rest to the engineers who own it.

  5. Architect

    Turn findings into durable information architecture — content that humans can navigate and machines can parse.

Exhibit BCurrent investigation

Machine learning on Amazon’s own silicon.

Open case · 2026–present

AWS Neuron SDK, Annapurna Labs

I document the software stack that runs PyTorch on AWS Trainium — for ML engineers who need models training on custom accelerators, not marketing copy.

The audience is expert and the stack is deep: framework integration, LLM serving, custom kernels, profiling. My job is making the working path unmistakable at every layer.

Subjects
Native PyTorch on Trainium · vLLM serving · profiling & observability tools · agentic development tools
Toolchain
Docs-as-code — Sphinx & RST, Git, CI, Read the Docs
Special angle
Designing documentation AI agents can consume: structured, front-loaded, machine-parseable
Exhibit CCase history

From financial risk models to voice assistants to ML silicon.

  • 2026 — present

    Senior Technical Writer, Amazon Annapurna Labs

    Cupertino, California

    AWS Neuron SDK documentation for Trainium.

  • 2024 — 2026

    Senior Technical Writer, Amazon Devices & Services

    San Jose, California

    Developer documentation for next-generation Alexa — authored with AI-assisted tooling before it was standard practice.

  • 2022 — 2024

    Programming Writer, Amazon Devices & Services

    Sunnyvale, California

    Alexa Skills Kit and Alexa Voice Service Device SDK docs; contributed to stealth generative-AI projects.

  • 2019 — 2022

    Senior Technical Writer, SS&C Algorithmics

    San Jose, California

    Financial-risk documentation. Led the move from DITA to docs-as-code (AsciiDoc, Git, GitHub) and the GitLab→GitHub CI/CD migration.

  • 2012 — 2019

    Senior Technical Writer, IBM

    Toronto & San Jose

    Risk analytics and aggregation docs for IBM Algo. Tooling support for quants authoring financial models in MathML; mentored interns through a FrameMaker→DITA conversion.

  • 2005 — 2011

    Technical Writer → Senior Technical Writer, Algorithmics

    Toronto, Canada

    User guides, configuration guides, and online help for financial risk management products.

Exhibit DField notes

Rules I work by.

  • Evidence over plausible prose.A sentence that sounds right and a sentence that is right are different sentences. I ship the second kind.
  • The working path is the product.Nobody reads documentation for pleasure. Every procedure gets executed, not just reviewed.
  • Navigation is architecture.Where a page lives — and what it sits beside — teaches as much as what it says.
  • Write for two readers.Humans skim; AI agents parse. Front-loaded, structured content serves both without compromise.
  • A summary is not proof.Trust what the system does, not what the ticket says it does.
Exhibit ECredentials

Applied math first, writing craft second, always learning.

Education

  • B.Sc. Applied MathematicsEast China Normal University
  • Technical Writing CertificateHumber College
  • Information Design CertificateUniversity of Toronto

Certifications

  • AWS Certified Machine Learning — AssociateEarly Adopter · 2024–2027
  • AWS Certified Solutions ArchitectAmazon Web Services
  • Statistics, HCI & DatabasesJohns Hopkins & Stanford, via Coursera
Off the record

Away from the case files: swimming laps, paper books, podcasts on long walks, and traveling for the next great meal. Home is South San Jose, shared with a hardware engineer and a budding technical writer — the family conversation runs from circuit boards to style guides.