AI Consulting

Since 2014 I built my reputation as a sought-after API technical writer in the financial technology and developer tools space. My work was precise, developer-first, and grounded in actual engineering workflows. Not just surface-level documentation. Companies in FinTech, e-commerce, and Web3 came to me when their API docs were holding back developer adoption, and they left with documentation that felt native to their platforms.

The work was always the same problem: engineers who had built something powerful but couldn’t explain it to the developers who needed it most.

Around 2021, something began to shift. The clients calling weren’t just asking for cleaner docs — they were asking harder questions. How should we structure our documentation as AI starts reading it alongside humans? How do we write for LLMs querying our API endpoints? What does developer onboarding look like when half the integrations are AI-generated? These weren’t documentation questions anymore. They were AI strategy questions. The shift was gradual, then accelerating. By the mid-2020s, I was as likely to be advising on ML workflow documentation, model inference pipelines, and AI-first onboarding strategy as I was writing a webhook guide or an OpenAPI spec.

Today I work at the intersection of technical writing, AI documentation strategy, and ML/AI upskilling — currently building hands-on experience with PyTorch, TensorFlow, and ONNX to stay ahead of the curve in a field where the docs and the models are increasingly inseparable. My LinkedIn profile https://www.linkedin.com/in/peterdavidgustafson.

Got an AI project? Get in touch with me.

Over the past 12 years I have consulted for some of the most recognized names in tech and finance — Coinbase, Epic Games, WEX Bank, and Discover Card. I have also worked with Interactive Brokers, Alchemy, ShipEngine, Stripe, and others. In 2022 I moved into AI consulting. I now integrate Claude, Gemini, and Copilot directly into documentation workflows. This cuts doc cycle times in half. It also reduces developer support tickets by up to 65%.

I specialize in REST and SOAP API docs, OpenAPI/Swagger specs, OAuth 2.0, and docs-as-code in Markdown. I build on platforms like Mintlify, Docusaurus, and ReadMe. Every project below is real work. Published pages and raw Markdown files are included.

— Peter Gustafson

Case Studies

Below are real-world solution-based projects I’ve worked on since 2022. Most of them were slow movers since some corporate security policies don’t play too well with LLMs. Thus, a few took time working SecOps and the software devs to create permission-based polices for full LLM integrations.

Consulted with Discover product owners and API engineers to upgrade partner-facing developer documentation for digital payments infrastructure. Documented token provisioning, cryptogram generation, and second-factor authentication flows end to end.

Used AI-assisted drafting to accelerate first-pass parameter definition tables, then validated output against live endpoint behavior. Applied prompt engineering to generate consistent request/response examples across multiple API surfaces.

Used LLM tooling to audit existing docs for terminology inconsistencies and flag gaps in error-handling coverage. Delivered documentation templates that their team could carry forward without ongoing consulting support.

This project focused on building hands-on fluency with Python-based ML frameworks to support AI/SDK documentation engagements. Documenting the full lifecycle from model training through export and deployment.

Writing sample guides for ONNX model export, inference pipeline setup, and runtime integration — the exact workflows AI developers need explained clearly. Using Claude and GPT-4 iteratively to pressure-test technical accuracy and surface edge cases in the documentation.

Building familiarity with transformer architectures and model serving patterns so I can write with genuine technical authority, not just surface-level paraphrasing of framework docs.

Advised the team on upgrading their TWS API documentation to best-in-class standards for a brokerage platform used by sophisticated developers and quantitative traders. Their engineers had decades of domain expertise but documentation that hadn’t kept pace.

Trained junior technical writers to formalize API docs into structured templates — transforming tribal knowledge into repeatable, scalable onboarding guides. Introduced AI-assisted review workflows so writers could validate technical claims faster without blocking on SME availability.

Applied AI tooling to identify structural inconsistencies across legacy docs and propose reorganization. The resulting templates now anchor all subsequent onboarding material without ongoing consulting involvement.

Coinbase hired me to update internal documentation and Salesforce architectural diagrams for new engineers at one of the world’s dominant crypto platforms. This engagement marked an early inflection point being my first client where AI-adjacent tooling was explicitly part of the conversation.

Rewrote API docs covering endpoint descriptions, authentication tutorials, and developer use cases. Used early AI writing assistants to generate first-draft descriptions from raw endpoint schemas, then refined for accuracy, tone, and developer empathy.

Added color and hierarchy to Salesforce diagrams to dramatically improve onboarding comprehension for new hires. Developed a repeatable diagramming approach that engineers could maintain without a dedicated technical writer on call.

Alchemy is one of the fastest-growing Web3 developer platforms to address a doc quality problem common to engineering-led companies: Swagger-generated output that was technically accurate but practically unusable by developers unfamiliar with the platform.

I reformulated choppy auto-generated doc structure into a clear narrative flow with tables, section blocks, and scannable formatting. Developed a simplified API doc template for their internal technical writers to apply consistently going forward all backtested by LLMs.

Used AI tooling to analyze existing docs for readability issues and generate before/after comparisons to demonstrate structural improvements to the product team. Also coded a call-to-action sign-up button embedded in their API overview page to convert developer readers into active users.

LLM Tools & Coding

Below are LLM tools and coding integration projects I’ve worked on. AI models change regularly so get in touch if you’re in need of an updated model for your next project.

Remember, LLM tools vary by algos and code base. Some are great for streamlining and automating while others are clunky. That’s why it’s critical to choose wisely when building an AI-first strategy.

GPT-3 / GPT-4 / GPT-4o | OpenAI

These are the models that kicked off the mainstream LLM era. GPT-3 proved large language models could write coherently; GPT-4 added reasoning depth and longer context. GPT-4o brought multimodal input and faster response. Widely used for first-draft generation, schema-to-prose, and API doc templating.

Claude Code | Anthropic

Anthropic’s agentic coding tool that operates directly in the terminal, reading and writing files, running tests, and navigating codebases autonomously. A step change for technical writers who work in docs-as-code environments — capable of making coordinated edits across large documentation repositories.

ChatGPT | OpenAI

The chat interface built on GPT models that brought conversational AI to a mass audience. Launched in Nov 2022, it became the go-to tool for iterative drafting, brainstorming, and content restructuring — particularly useful for working through documentation outlines and tone revisions in real time.

Claude (Sonnet / Opus) | Anthropic

Anthropic’s flagship model family, built with a strong emphasis on safety, long-context reasoning, and careful instruction-following. Claude excels at technical writing tasks, nuanced document editing, and extended back-and-forth refinement — making it a natural fit for complex API documentation work.

Gemini (Pro / Ultra) | Google DeepMind

Google’s multimodal LLM family, integrated across Workspace and developer tooling. Gemini’s strength is its native integration with Google Docs, Drive, and Search — useful for documentation workflows that live inside enterprise Google environments or require real-time web grounding.

GitHub Copilot | OpenAI

AI pair programmer embedded directly in code editors. Beyond code generation, Copilot is increasingly used to write inline code comments, docstrings, and README content alongside working code — blurring the line between developer tooling and technical writing assistance.

Llama 2 / Llama 3 | Meta AI

Meta’s open-weight model series, freely available for commercial use. Llama models enabled teams to run capable LLMs locally or in private cloud environments. It’s user-friendly for documentation work involving sensitive financial or proprietary API data that can’t leave the organization.

Mistral / Mixtral | Mistral AI

A European open-weight LLM known for punching above its weight class at smaller parameter counts. Mixtral’s sparse mixture-of-experts architecture delivers fast, cost-efficient inference — popular for self-hosted documentation pipelines and automated doc generation workflows.

Perplexity AI | Perplexity

A search-native AI that answers questions with cited, real-time web sources. Especially useful during documentation research phases that quickly models current API behavior, changelogs, and developer forum context without manually hunting across multiple sources.

o1 / o3 | OpenAI

OpenAI’s reasoning-focused model series, designed to think through complex problems step by step before responding. Particularly valuable for validating technical accuracy in API docs — catching logical gaps in endpoint descriptions, authentication flows, and error-handling sequences.

Cursor | Anthropic

An AI-native code editor built on top of VS Code that uses Claude and GPT models to assist with code and prose simultaneously. Useful for docs-as-code workflows where Markdown documentation and code samples live in the same repository and need to stay in sync.

NotebookLM | Google

A research and note-taking tool that lets you upload source documents and query them conversationally. Valuable for onboarding into complex API ecosystems quickly — upload existing docs, SDKs, and changelogs, then interrogate them to understand the full surface area before writing.