Software Engineer · Optimization Systems

Building systems that
schedule the impossible.

I'm Steve Duncan — I design AI native enterprise software, from full-stack ERP platforms to Optimere, a medical scheduling app. Right now I'm exploring quantum approximate optimization (QAOA) to solve the rescheduling problems classical systems struggle with at scale.

Run the QAOA demo → See Optimere
MEDICAL SCHEDULING
QAOA
QUANTUM R&D
01 / PRODUCT

Optimere

A medical scheduling platform that keeps clinics running when reality doesn't cooperate. Optimere coordinates providers, rooms, equipment and patients in real time — and when a cancellation or no-show ripples through the day, it rebuilds the schedule around the disruption.

Built on a full-stack engineering foundation using modern AI development methods: complex domain modeling, hard constraints, and operations teams who need answers in seconds, not minutes.

React Node / TypeScript PostgreSQL Constraint solver
Today · Halifax Medical Clinic
optimized
DR. PATEL
DR. OKafor
DR. LEE
8:00
Intake · M. Cole
rm 2 · 30m
Follow-up
rm 4 · 20m
open
8:30
open
Imaging · R. Diaz
rm 1 · 45m
Consult
rm 3 · 25m
9:00
No-show ⚠
rebalancing…
Intake · J. Park
rm 2 · 30m
Follow-up
rm 4 · 20m
3 conflicts resolved · 1 pending utilization 92%
Live rebalancing
Cancellations trigger an automatic re-solve across providers and rooms.
Hard constraints
Equipment, credentials, and patient prep windows are never violated.
02 / RESEARCH

Quantum-assisted rescheduling

Rescheduling is a combinatorial optimization problem — the kind that explodes as a clinic grows. I'm prototyping QAOA (the Quantum Approximate Optimization Algorithm) as a future engine for Optimere.

Today the math is small enough to run classically, so I use an AI-driven classical simulation of the quantum circuit to design and validate the approach. The architecture is built to hand off to real quantum hardware the moment it becomes commercially viable at the scale of large patient datasets. The demo below is that simulator, running live.

LAYERS p
iter 0
conflicts left
idle
Conflict graph
appointments partitioned into two slots · cut = resolved
Cost convergence
⟨C⟩ expected resolved conflicts per iteration
⟨C⟩ / —
Measurement spectrum
probability across 2ⁿ candidate schedules · best highlighted
Parameter landscape
⟨C⟩ over (γ, β) · marker = current params
γ β
Max-Cut formulation · 6 appointments · classical state-vector simulation gradient-ascent over QAOA angles