The problem we are paid to solve.
Your Fortune 500 engineering team runs tens of thousands of jobs every month on classical HPC. Some of them bleed millions. Some will benefit from quantum acceleration when the hardware matures.
Nobody can tell you which. That is the audit.
Two weeks. Four steps. One report.
Kickoff & Scope
Mutual NDA. Scoping call. We provide the GPG-signed CLI binary (Linux x86-64, glibc 2.28 or newer; see FAQ). Your team confirms which Slurm / K8s / PBS clusters are in scope.
On-Prem Audit
Your team verifies the binary, supplies your compute rates and an optional job-metadata file, and runs the CLI against scheduler logs on the head node. Without the metadata file the run scores cost only and cannot produce strong archetype matches. No raw data leaves your network.
Review & Consent
The PDF report is generated locally and stays with you. If you choose to contribute to the Latency Index, you preview the sanitized manifest and type the consent token; the CLI never uploads it — any transfer is a manual step you perform.
Debrief
Two-hour debrief with your engineering team on the report you choose to share with us. First-cohort reports carry no advisor signature (disclosed in the SOW).
The 15 archetypes.
Every workload in your audit is classified against this library. Each archetype carries a hardware horizon and a confidence band, both published in the methodology.
Optimization
Combinatorial QAOA
Max-Cut / MIS / scheduling
Portfolio QAOA
Allocation with budget + risk constraints
MILP-mapped routing
Vehicle routing / network flow
Simulation
VQE molecular Hamiltonian
Drug discovery, small-molecule energetics
Quantum chemistry
Electronic structure beyond DFT
QPE eigenvalue
High-precision spectral problems
Search
Grover unstructured
Pre-image / SAT search
Grover-mapped CSP
Constraint satisfaction
Linear Algebra
HHL
Sparse linear systems (caveats apply)
Amplitude Estimation
Monte Carlo / derivatives / risk
Number Theory
Shor's algorithm
Factoring / discrete log
ML / Hybrid
Quantum kernel methods
SVM / classification on quantum features
Variational gradient
Quantum-assisted training
Graph / Stochastic
Quantum walks
Graph traversal / centrality
Stochastic optimization
Optimization under uncertainty
What this engagement actually buys you.
Three columns. The third column is the moat.
What we measure
- Every job classified against the 15-archetype library with high / medium / low / speculative confidence
- QSS computed per workload, five inputs, five weights, fully auditable
- Annualized classical cost per workload, from your supplied compute rates and observed runtime, printed in the report
- Queue-wait cost for the top 10 workloads, reported separately and never added to the annualized cost
- Hardware horizon band per archetype (2026 / 2028 / 2030+ / speculative), as assigned in the methodology
What we deliver
- Board-ready PDF: executive summary, per-workload table (hashed id, archetype, match strength, QSS, band, annualized classical cost), the top 10 workloads by cost, the inputs used, and the claims the report refuses to make
- Methodology version and hash in every page footer, pinned to the published methodology
- Reproducible results: the same logs, rates, key, and methodology version produce a byte-identical report
- Two-hour debrief call with your engineering team
What we won't claim
- Specific speedup multipliers on today's NISQ-era hardware
- Quantum-advantage delivery dates we cannot defend with published evidence
- Vendor performance comparisons (IBM vs. IonQ vs. neutral-atom vs. anyone)
- ROI projections beyond classical bottleneck cost on your own infrastructure
- Forced classifications when no archetype credibly fits
This is the page no consultant writes. It is the page Fortune 500 CTOs pay for.
The Quantum Suitability Score, in practice.
Real output from the CLI's bundled sample logs: the same five inputs, two strong matches, one honest no.
Sample job: VQE molecular Hamiltonian (strong match)
High: design-partner candidate when the stated-horizon hardware ships. Track closely.
Sample job: portfolio optimization (strong QAOA match)
High on structure and horizon. Classical Cost is 1 because the sample spend sits near the $1k/yr floor.
Sample job: gradient boosting on tabular data (no reduction)
Not a candidate: classical is the right answer.
Scores above come from `qintercept audit --sample` (synthetic logs and metadata, CLI v0.0.4). Real scores are computed deterministically from your scheduler logs, your compute rates, and your metadata file, and are reproducible from those inputs.
Runs inside your firewall. Period.
The audit CLI is a single GPG-signed binary. It runs on your HPC head node and reads scheduler logs and the metadata file you supply, locally. It never opens a network connection.
The only artifact designed to leave your network is the sanitized JSON manifest, and only if you write it with the consent token and move it yourself: keyed-HMAC workload ids, archetype / band / cost-band labels and counts, scheduler type, methodology version. No raw logs. No file paths. No user IDs. No dollar figures. No source code.
What lands on your desk on Day 14.
A board-ready PDF, generated on your head node. Methodology hash on every page. No advisor signature until an advisor is seated.
One SKU. Forty thousand dollars.
Single SKU for the first ten engagements. Tiered pricing after. The first design-partner cohort runs on per-engagement terms without an advisor signature, disclosed in writing before contracting.
Included
- CLI license + on-prem deployment support
- Board-ready PDF report
- Anonymized Latency Index ranking, once the index publishes (after the first five paid audits)
- Two-hour debrief call
- Public methodology (1.0.0-draft), built for advisor sign-off
Not included
- Ongoing monitoring (scheduler product, on the roadmap)
- Pilot implementation work (separate SOW)
- Custom archetype development (case-by-case)
- LSF support in v1