Eight chapters: an oscillator PINN, combinatorial bit encodings, variational circuit plumbing, a hybrid classifier, finance and routing QUBOs, tensor-train LLM compression, and deployment governance.
Read in chapter order or open the chapter that matches your current bottleneck: residual versus boundary trade-offs, honest baselines for a quantum layer, penalty tuning on a QUBO, or evidence packaging for a release review.
The open repository sulimovp/sds_quantum_workshop holds runnable workshop code and installation notes; Implementation sections link there when you need the full cells.
Start from a small ODE you can audit: boundary data, collocation grid, and a residual that must vanish if the model is faithful.
Read Story →Many classical hard problems become searches over structured bit assignments; quantum hardware reads bitstrings after measurement, so the encoding step is never optional.
Read Story →This chapter is the skills bridge: exact statevector checks, hardware-aware patterns, and the classical outer loop that every VQE or QAOA deployment shares.
Read Story →Same dataset splits, same loss, same reporting: the quantum block must earn its place against a classical baseline you would ship.
Read Story →Each bit is an invest-or-skip decision; the QUBO encodes return appetite, risk via covariance, and hard budget constraints through penalties.
Read Story →One-hot encodings make constraints explicit: each city once, each position once, distance on legal tours only.
Read Story →Matrix product state structure reappears as tensor train cores: fewer parameters, controlled error, no cryogenic requirement.
Read Story →Translate working science into decisions procurement, legal, and SRE teams can defend: parity, reproducibility, runtime envelopes, and explicit classical fallback.
Read Story →