1UP TRENT
SCORE 00000000
PRESS START
CREDIT 01 RESUME

PRESS START · A PORTFOLIO

TRENT McFARLANE

AI ENGINEER · GAME DEV

I build AI agents for enterprise clients at Insight, after building AI at Microsoft and Alaska Airlines. On the side, I make games. So I turned my career into a short film that plays like one: four levels, two side quests and a boss battle.

PLAYS THE 3:39 FILM SOUND ON

LEVEL SELECT

  1. LEVEL 1SELL
  2. LEVEL 2OPERATE
  3. SIDE QUESTINBOX
  4. BOSS BATTLESUPPORT BOT
  5. LEVEL 3SCALE
  6. LEVEL 4LEVEL UP
  7. INVENTORYGEAR
  8. BONUS ROUNDGAMES
PLAYER 1 · CHARACTER SELECT

Player 1: Trent McFarlane

Trent McFarlane
TRENT McFARLANE
P1
CLASS
AI ENGINEER
CLOUD INFRADEV PLATFORMSAI PRODUCTS
GUILDS
FORMERMICROSOFTFORMERALASKA AIRLINESNOWINSIGHT
STATS
  • CLOUD SAVINGS$3M+
  • KEYNOTE CROWD500+ 2022
  • LIVE PRODUCTS8 TODAY
  • PATENTFILED
SUB-CLASS
GAME DEV ★
LEVEL 1 · SELL

Agents that help people sell

At Microsoft, I led the team that built SalesCoach, where sellers practice live conversations with an AI customer, with speech recognition, speech synthesis and an interactive avatar.

At Insight, for a Fortune 500 energy company, an agent starts the moment a deal opens in Salesforce, finds and ranks the right stakeholders, and writes them back to the CRM.

PILOT: 50 REPS · BUILT FOR 1,500+

It's the first of a fleet. A Market Watchdog is underway, with Competitor Watchdog and Business Analyst agents planned.

SIDE QUEST · MYSTIC SHOPPE

And for a real shop, Mystic Shoppe, I launched its site, bookings and inventory in 60 days. Plus an AI that reads their monthly calendar and drafts every event for them to approve.

LAUNCHED IN 60 DAYS

The Stakeholder Profiler from the film: a new Salesforce opportunity, its ranked stakeholders, and the agent writing them back
Client work is redrawn for the film, with names redacted.
Mystic Shoppe: the shop's monthly calendar read by an AI that drafts each event for the owners to approve
LEVEL 2 · OPERATE

Every incident, investigated by an agent first

At Alaska Airlines, cloud incidents took 3 days to 2 weeks to resolve. My team built an agent that investigates each one first and finds the closest past fix.

3 DAYS–2 WEEKS → MINUTES–A FEW DAYS

PILOTED Predicting incidents before they happen.

Then I rebuilt the pattern solo, in the open: a support triage agent for HackerRank's 24-hour Orchestrate hackathon. It took 8th place out of 1,349 teams.

8TH PLACE / 1,349

An incident card linked to a past fix, with the time to resolve now minutes to a few days
The hackathon high-score board: Trent, solo, in 8th place of 1,349
SIDE QUEST · INBOX

A reply for every recruiter

Post "open to work" as an AI engineer, and messages arrive faster than anyone can answer them well. So I built ReplyGuy.

It's a Chrome extension that remembers what I'm looking for, my resume, how I write, and every conversation I've had with each recruiter. Then it drafts replies in my voice, right inside LinkedIn and Gmail, calendar link included.

5 DRAFTS · IN YOUR VOICE

ReplyGuy's memory (goal, resume, writing style, contact history) and five drafted replies, each with a calendar link
The inbox is a lookalike drawn for the film.
BOSS BATTLE · YOUR AI AGENT VS. THE SUPPORT BOT

So I sent in my own AI

When my internet went down, my provider's support bot wouldn't let me reach a human. So I sent in an agent with browser use.

It got a person on the line, booked a same-day repair, checked the 2–5 PM window with me first, and got me a refund.

K.O. · REFUND ISSUED

K.O.: the agent beats the support bot, refund issued
The support chat in the film is recreated.
LEVEL 3 · SCALE

I cleaned up Alaska's AI and saved $2M

Alaska Airlines brought me in to clean up its AI infrastructure, so I founded the AI Operations Program. We reviewed 30 projects in 2 months, cut 60 AI instances to 25, and folded costly projects into one small team.

Then we rebuilt what was worth keeping on one agent platform. Teams secure their data once, hand it to agents, and reuse Alaska's UI to ship faster.

  • $2Msaved
  • 60 → 25AI instances
  • 2 WEEKS → 2 DAYSto get AI access
  • 100+teams building on the platform
Level 3 from the film: Alaska's AI cleanup and the $2M it saved
LEVEL 4 · LEVEL UP

Training plans in under two weeks, and labs that grade themselves

At Microsoft, a training plan in Azure and Microsoft's products, for 2,000 to 10,000+ of a customer's employees, took 3–4 months. With Skill Plan Assist, now Microsoft's AI Skills Navigator, it was ready in under two weeks.

3–4 MONTHS → UNDER 2 WEEKS

I also built the experience that launches and scores hands-on labs inside Microsoft courses, so certification went from 48 hours to instant.

48 HRS → INSTANT

The training-plan quest: Azure and Microsoft products for 2,000 to 10,000+ employees
INVENTORY

Every run needs gear, so I build my own

Two MCP wiring diagrams: Claude Code to a LinkedIn MCP adapter to the LinkedIn API, and an agent to PostQueue to Buffer and Reddit

MCP SERVERS

My MCP servers plug those agents into LinkedIn and my social posting queue: Claude Code to a LinkedIn MCP adapter to the LinkedIn API, and an agent to PostQueue to Buffer and Reddit.

BONUS ROUND · GAMES BY TRENT

And yes, I make games

DEEPWELL is my deckbuilding roguelike: five heroes descending a flooded god-shaft. It's in playtest now, with early-access signups open.

  • DUSTLINEDirt-bike trials · ride it ↗
  • CRAWLERTwin-stick dungeon shooter
  • NEON FELTMultiplayer poker
  • STORY RPGSAI story games
  • CLIENT WORKA fighting game for a client
DEEPWELL gameplay: a hand of cards against the Sump Colossus
A shelf of Trent's games: CRAWLER, NEON FELT, DEEPWELL, DUSTLINE, AI story RPGs and a fighting game for a client

CONTINUE?

Let's build something.

FINAL SCORE 00000000