MartianAI co-founding engineer, Head of Europe
Architected the AI model router, with 1000x faster inference. Built and led a team of 6, helping grow company value by tens of millions.
What changes for your team
More out of the AI you already use.
Deliver more product
A better workflow, so the same team ships more, faster.
Agents that get it right
Better context and setup, so agents do good work more often.
Reviews that keep up
Review moves as fast as your team writes code.
Quality keeps up with speed
More output without more bugs reaching production.
Gains you haven't found yet
AI working in places your team hasn't tried.
Best practices, team wide
Every engineer works the way your best one does.
What we explore together
Where there's more to gain.
How your team uses AI
Who uses agents, for what, and where there's more to gain.
Agent setup
Whatever harness you use: CLAUDE.md, AGENTS.md, skills and access, so agents have the context they need.
PR reviews
How AI-written code gets reviewed, so reviews keep up with output.
Workflow and gates
The steps, gates and sandboxes that let agents do more, safely.
AI coverage
Places where AI could help but isn't used yet.
Cross-team learnings
What one team figured out, shared with the rest.
I've built this inside real teams. At Reddy, I built AI agents that investigate bugs, review code and ship fixes. Let's find what fits yours.
How it works
Two weeks, then your call.
- Week 1
Discover
We look at how your team works with AI today, with read access and short talks with a few of your engineers.
- Week 2
Quick wins
We set up easy fixes and process improvements, and agree on a roadmap for the rest.
- After
Keep going
Work through the roadmap with me, or run it yourselves. Your call.
What you get
Results in two weeks, not a slide deck.
- Each area scored, with evidence
- Quick wins and process fixes, set up within the two weeks
- A suggested roadmap for what comes next
- A 1 hour walkthrough with your leads
Typical quick winsExamples
- Shared agent rules
- One AGENTS.md or CLAUDE.md for your main repo, so every agent starts with the same context.
- Safe access
- Agents can read logs and staging, without touching production.
- Plan first
- A short planning step before agents write code.
- AI PR review
- A review checklist for AI-written pull requests.
- Sandboxed agents
- Agents run in a sandbox, so nobody has to watch every session.
I'm Dory.
An AI engineer who builds these systems.
“There are very few people in the world who can take a digital product from concept to working software without much day-to-day direction, let alone do it quickly. Dory, Coreteam lead engineer can, and is great to work with along the way.”
ReddySenior staff AI engineer
Made a voice AI platform 10-15x faster and cut voice processing costs 5x. Built AI agents that find bugs, review code and ship fixes.
BCGTech lead, CO2 AI
Took the platform from MVP to enterprise product. Cut customer onboarding from hours to minutes.
HearstSenior engineer
Took over and rebuilt the systems behind a $100M advertising business, then hired the team that took over.
I hit the ground running. I've taken over existing systems many times, often leading within the first two weeks. My last two clients both hired me directly.
Full background on LinkedInMy AI research: Karpathy vs 8 CLAUDE.MDs (126 builds) and which planning method works best with AI (50 builds).
