Senior Software Engineer Sydney, Australia Open to senior engineering roles
I follow the product requirement as deep as it needs to go.
Senior product engineer. I take ambiguous product problems from requirement to production, across the frontend, backend, data and AI layers needed to make them work. When the straightforward solution stops holding, I trace the problem into the underlying system and build the fix. Previously Atlassian Growth; now building independent AI products.
Two systems, one method
01 / read acrossread down, twice
The same four questions, asked of both systems.
CH 01
Codenames AI
CH 02
Experiment measurement
01
Requirement
What the product had to do, before any mechanism existed.
Players need an AI teammate whose every move is legal on this board, unsupervised, in a product that is live.
Growth teams had to know whether a shipped change actually worked — including for an acquired product that was not on experimentation infrastructure yet.
02
Depth required
The layer that had to be opened when the assumption underneath stopped holding.
A model proposes a clue. Nothing in the reply guarantees it is legal on this board. Schema-first structured outputs plus board-aware domain validators separate valid JSON from a legal move. Model migrations are treated as controlled experiments, with evaluation built into product behavior.
Attribution windows and pipeline gaps silently change what an in-flight experiment appears to say. A Cross Flow funnel observability audit and an authored attribution formula adopted for Growth Experiment Impact Estimation, plus embedded event-pipeline work that restored attribution for in-flight experiments.
03
Contract
What the system is now allowed to claim.
Valid JSON is not a legal move — the validator decides.
Preserved statistical validity without restarting experiments.
04
Qualified evidence
One number, inseparable from the scope it was measured in.
175+
monthly active players
Live-product telemetry · resume phrasing uses 175+ as durable floor, not the rolling monthly-active-players snapshot
9%–41%
uplift variance across StatSig attribution windows
In-flight experiments · Statsig · the range is the finding
01
Requirement
What the product had to do, before any mechanism existed.
02
Depth required
The layer that had to be opened when the assumption underneath stopped holding.
03
Contract
What the system is now allowed to claim.
04
Qualified evidence
One number, inseparable from the scope it was measured in.
System 01 of 02
CH 01
Codenames AI
01Requirement
Players need an AI teammate whose every move is legal on this board, unsupervised, in a product that is live.
02Depth required
A model proposes a clue. Nothing in the reply guarantees it is legal on this board. Schema-first structured outputs plus board-aware domain validators separate valid JSON from a legal move. Model migrations are treated as controlled experiments, with evaluation built into product behavior.
03Contract
Valid JSON is not a legal move — the validator decides.
04Qualified evidence
175+
monthly active players
Live-product telemetry · resume phrasing uses 175+ as durable floor, not the rolling monthly-active-players snapshot
System 02 of 02
CH 02
Experiment measurement
01Requirement
Growth teams had to know whether a shipped change actually worked — including for an acquired product that was not on experimentation infrastructure yet.
02Depth required
Attribution windows and pipeline gaps silently change what an in-flight experiment appears to say. A Cross Flow funnel observability audit and an authored attribution formula adopted for Growth Experiment Impact Estimation, plus embedded event-pipeline work that restored attribution for in-flight experiments.
03Contract
Preserved statistical validity without restarting experiments.
04Qualified evidence
9%–41%
uplift variance across StatSig attribution windows
In-flight experiments · Statsig · the range is the finding
One practice
Not a pivot — the same method, two decades.
Atlassian, 2014–2025 — Software Developer, Engineering Manager, Senior Software Engineer. Independent since 2026.
Four of those years were spent managing, and the work was recognisably the same. The mechanisms each era left behind — attribution formulas, experiment-operations rules, checks that run at the point of change — are the ones running in the independent work now.
Practice record, era by era →8–10
engineers directly managed
Atlassian Growth organization · through significant organizational and product change
Open to senior engineering roles — michael@multipliers.dev