Active — 185 tests passing, MP5 next

qsim-studio

A quantum computing simulator built from first principles — no Qiskit, no Cirq in the core — with a visualization layer for every milestone.

Quantum ComputingGPU Computing

Inside the work

A state, transformed.

Concept in motion
|0⟩|1⟩PURE QUBIT / Rᵧ(θ) ROTATION

An illustrative pure single-qubit state on a Bloch sphere. The state vector rotates on the sphere surface; this is not a running quantum computation.

01

Prepare

Start with a statevector and a circuit expressed through the simulator’s gate representation.

02

Apply a gate

A single-qubit unitary changes the state; for a pure qubit this can be visualized as a rotation of the Bloch vector on the sphere surface.

03

Measure

Read probabilities from amplitudes and compare the result with the expected circuit behavior.

qsim-studio is a local, GPU-accelerated quantum simulator built module by module against a public roadmap, with the explicit rule that external quantum frameworks (Qiskit, QuTiP) are only ever used inside an isolated validation layer to cross-check the from-scratch implementation — never inside the engine itself. Everything is pure Python plus NumPy/CuPy.

The build runs milestone-by-milestone: a quantum math engine (linear algebra, gates, single-qubit Bloch representation), a multi-qubit statevector register using tensor-contraction gate application, a measurement and statistics layer (Born-rule sampling, full/partial collapse, expectation values), and a circuit layer with its own intermediate representation, a QASM-2-subset parser/exporter, a gate-fusion optimizer, and an SVG circuit-diagram renderer.

Tests passing per completed milestone

021.5443.0764.6186.14Tests passing — MP1: 2424MP1Tests passing — MP2: 3636MP2Tests passing — MP3: 5252MP3Tests passing — MP4: 7373MP4

Roadmap milestones

MilestoneScopeStatusTests
MP1 — Quantum math engineLinear algebra, gates, single-qubit Bloch sphereDone24 pass
MP2 — Multi-qubit statevectorTensor-contraction gate application, Bell/GHZ states, entanglement entropyDone36 pass
MP3 — Measurement & statisticsBorn sampling, collapse, expectation values, basis rotationDone52 pass
MP4 — Circuit layerIR, QASM-2 subset, gate-fusion optimizer, SVG rendererDone73 pass
MP5 — Foundational algorithmsNext milestonePlanned—

Key Findings

What the work shows

Every milestone ships its own visualization: a Bloch sphere driven by real ⟨X⟩⟨Y⟩⟨Z⟩ expectation values, GHZ-state amplitude bars with entanglement entropy, an exact-vs-shot-noise measurement histogram showing convergence, and a circuit diagram rendered directly from the IR — not hand-drawn.

MP2's tensor-contraction statevector engine is cross-checked against a dense-matrix ground truth, and MP1's gate math includes oracle tests that activate automatically once Qiskit/QuTiP are installed, without becoming a hard dependency.

185 tests pass across the four completed milestones (24 + 36 + 52 + 73) with zero external quantum-library dependency in the core engine.

PythonNumPyCuPy