Forrest Newman

Staff Software Engineer · AI & Product Engineering

I turn slow, high-stakes manual work into AI systems that ship, and I prove they work.

I'm the lead engineer for Prime Video's AI-native Trust & Safety platform, where AI agents and human reviewers decide whether a title is released or blocked worldwide. Over 13 years, 9 of them at Amazon, I've owned both the architecture and the product direction, usually without a dedicated PM, designer, or architect.

Open to Staff and Principal roles with product ownership · Remote

55% less handling time for reviewers in a production trial of the AI agent platform I lead
240 operators' worth of work saved every year by review tooling I led
26 → 2 weeks to launch Amazon checkout fraud protection in a new country
80% lower infrastructure cost from a database redesign I led

What I bring to your team

AI that holds up in production

I treat agents like any critical system: engineered context, layered guardrails, challenge agents that check the work, and evaluation against real human decisions on production traffic instead of a test set the prompt was tuned to. In trial, the AI agreed with human reviewers 98.73% of the time.

Product judgment, not just tickets

When there's no PM, I go to the people doing the work. I've shadowed reviewers in Seattle, Austin, and Costa Rica, written requirements with program teams, and built working prototypes so leaders could decide with something real in front of them.

Scale, reliability, and cost

I've run checkout systems at 10,000+ TPS across 10+ countries and taken a critical API from 0.45 to 83.6 TPS ahead of re:Invent. I migrate without drama: parallel runs, feature flags, zero downtime, and bills that go down.

Leadership that multiplies

I've led up to three teams at once (30 engineers) across 5+ orgs, and I work directly with directors and present to VPs. I mentor about three people at a time. Mentees include 5+ engineers promoted, two promoted to Senior, and a content reviewer who moved into a technical role.

Selected work

Each started as a problem real people were struggling with. Here's what I did and what changed.

Prime Video Trust & Safety · 2026

AI agents that do the legwork, with humans making the call

Problem
Reviewers gathered evidence by hand from many systems to decide on ratings and policy violations, and every change to a review process needed an engineer.
What I did
Led the design of an end-to-end platform on AWS Bedrock and Step Functions. Agents gather evidence and write a report, challenge agents verify it, and a reviewer makes the final decision. Program managers configure each workflow's prompt, guardrails, tools, model, and review screen themselves, with no engineering request.
  • 55% less handling time
  • 98.73% agreement with human decisions
  • 100% match with the legacy tool's outcomes

Trial run on shadow production traffic. Launching to 200 operators in December 2026.

Prime Video Trust & Safety · 2025

One screen instead of nine systems

Problem
A single review decision meant pulling information from 9 different systems, and new countries were coming with no tooling ready.
What I did
Watched operators work on site, then shipped a review tool with every signal on one screen, model output inline, guided decisions, and automatic task assignment. I then migrated and consolidated the remaining workflows so a legacy tool, where every update took two months, could be retired.
  • 75% less manual review time
  • 240 operators' work saved per year

Prime Video Partner Experience · 2024

Partner Action Center

Problem
Studios delivering content couldn't see what was expected, what had arrived, or what had failed. The answers were spread across 5+ orgs, and feedback could take a week.
What I did
Gathered requirements directly with Amazon MGM Studios, got every org to contribute its signals, and designed the DynamoDB single-table service behind the portal. My design docs were detailed enough that the team kept delivering on plan while I was away.
  • Feedback in 12 hours instead of 7 days
  • 42% faster issue resolution
  • 12% better on-time delivery
  • 16.5% higher usability score

Serves 3,400+ partners and 30M+ title parts.

Amazon Retail Secure Delivery · 2019–2022

Fraud protection at checkout scale, in any country

Problem
Deciding which packages need a signature or one-time passcode relied on country logic hard-coded throughout the system. Launching a new country took 26 weeks.
What I did
Led 8 engineers in rebuilding it as a configuration-driven global system with per-country ML models and rules. We launched changes through A/B tests and streamed every decision to the data lake.
  • New country in 2 weeks, down from 26
  • 10,000+ TPS across 10+ countries

This work earned my promotion to Staff (SDE III).

More results

Built with AI, end to end

Private Chef: a business site the owner runs alone

Side project · January 2026 · Idea, design, and build with Claude

Live demo

I wanted to find out how far I could take a real product with AI as my partner at every step. I came up with the idea, designed it, and built it with Claude from start to finish for a friend's private-chef business in Mexico.

  • The owner never needs a developer. Every line of text, every image, and every product can be edited right on the site from the admin side.
  • Built for the real customer. It's fast on phones and slow connections, and it costs very little to run.
  • Production-ready on AWS. Serverless, secure admin sign-in, a custom domain, and order emails, all deployed from scripts.
Request admin password

The demo uses sample data and images. Request access below and I'll send you a personal password that opens both the admin site and the code.

Experience

  1. 2020 – Present

    Staff Software Engineer (SDE III) · Amazon

    Technical lead across Prime Video: Trust & Safety (AI agents, policy detection, reviewer tooling), Partner Experience, and Partner Quality. Before that, Amazon Retail Secure Delivery.

  2. 2017 – 2020

    Software Development Engineer II · Amazon

    Secure Delivery at checkout, the AWS Training site, and AWS Marketplace Seller Billing. Stepped up to lead teams in each.

  3. 2013 – 2017

    Lead C# API Developer · Prolifiq Software

    Led a 6-engineer backend team building the APIs for a SaaS platform. Improved data-layer performance 90% and owned incident response.

  4. 2008 – 2013

    Customer Support Engineer II · Prolifiq Software, Selectron Technologies

    Where I started: debugging customer problems, building client features in C# and SQL, and rewriting IVR phone-system engines.

Toolkit

AI & product

  • AWS Bedrock agents
  • Guardrails
  • MCP
  • Context engineering
  • LLM evaluation & shadow testing
  • Human-in-the-loop
  • Claude Code
  • Kiro
  • UX research

Cloud & architecture

  • Step Functions
  • Lambda
  • DynamoDB
  • S3
  • API Gateway
  • ECS
  • Kinesis
  • Event-driven systems
  • Microservices

Languages

  • Java
  • TypeScript
  • React
  • C#
  • Python
  • SQL

B.S., Information Technology (Application Development), Oregon Institute of Technology

Let's talk

I'm looking for a Staff or Principal role where I own the architecture and help shape the product, especially in AI and human-in-the-loop systems. Send a note and I'll reply personally.

View the code

Enter the personal password I sent you to get a private, read-only link to the code. The link works for 24 hours.

No password yet? Request access.