A systematic quality-control backstop for transportation engineering plan sets. Hundreds of engineering rules from your agency's published standards, applied uniformly to every sheet you ship. Caltrans live today — built to extend to any agency or firm style.
Drop us a line — we'll respond within a few business days about timing and pilot terms.
A 20-second look at the actual flow.
Plan-set quality control is the systematic pass through every drafting and design rule before submittal — checking that each sheet conforms to the standards your agency or firm requires. Done by hand, it's tedious, error-prone, and the first thing to slip under deadline pressure. Stratum Check runs that pass automatically, surfaces what it finds, and routes everything to your engineer for the call.
Plans Preparation Manual — title blocks, sheet layout, annotations, submittal requirements.
Line weights, symbology, layer conventions, and plot-visible drawing standards.
Highway Design Manual — alignment, sight distance, pavement, structures. Premium tier, in development.
Cross-reference QA between the plan set and Caltrans standard special provisions.
The architecture is built to extend — custom rule sets per agency, firm-specific style overrides, and AI-assisted rule creation as we onboard new standards.
No other state DOT has produced drafting and design standards as rigorously codified as the Caltrans PPM, CADD Manual, and HDM. Every sheet in every Caltrans PS&E follows a published rule. That makes it the cleanest possible starting point for a systematic QC engine — the rules exist on paper, the conventions are stable, and the cost of getting them wrong at submittal is concrete.
Other transportation agencies are on the roadmap. The framework generalizes to any agency that publishes its own standards. If your firm delivers across multiple state DOTs and you want to shape how Stratum Check generalizes, get in touch.
Three passes across every sheet — text, geometry, and vision — applied to every rule that fits.
Drop the full plan set PDF. Stratum Check extracts text and rasterizes every sheet at the resolution the engine needs.
Title, typical section, layout, profile — each sheet is classified from its title block and filename so rules apply where they belong.
Each rule is hand-built by engineers from published agency standards. Deterministic checks run first — text extraction and geometric verification. AI vision checks only where pixels are the only path to the answer.
Your reviewer confirms, defers, or dismisses each finding with a single click. Citations included. Engineer keeps the call.
Reviews in flight, recently completed, and how your credit balance is tracking. Click any review to see findings, work through them, or check the engine log.
The archive. Drill into any past run to confirm or dismiss findings, export the report, or compare against the next revision.
Sheet count, severity-graded findings, engine version, finish time. Click any finding to dive in.
Anyone who has delivered a Caltrans plan set under a real schedule knows the pattern. The engineering is sound. The drafting is competent. The QC manager is sharp. And then the package goes to district, and the comments come back — postmiles missing on three sheets, line weights inconsistent on the layouts, a sheet index that drifted from the standard placement, an annotation that should have been caught.
Not because anyone was careless. Because a manual pass through hundreds of published rules across 70-plus sheets, in the last 48 hours before submittal, is something humans simply cannot do reliably. The result is rework, re-stamps, schedule slip, and professional exposure that nobody on the project actually deserved.
Stratum Check is the systematic pass that doesn't depend on human stamina. It runs every applicable rule on every sheet, every time, with citations back to the source. Then your engineer reviews what it found and makes the call.
Stratum Check is built for the AI era — but it isn't an LLM with a stack of plans dropped in. The rules are designed by engineers and run deterministically. AI handles the checks that genuinely require vision. Both are deliberate. Both are auditable.
Every rule is engineer-designed and validated before going live, drawn from published agency standards. Rules run as deterministic code — same input, same output, every run. AI may help engineers draft candidates from new standards documents, but the runtime engine never asks a model what "the rule should be."
Two places it runs: classifying each sheet's type so the right rules apply, and performing vision-pass checks where pixels are the only meaningful signal. Everywhere else, the engine is deterministic code. We don't ask a model "is this plan set OK?" — we ask it precise, scoped questions tied to specific rules.
Each finding traces to a specific check, the standard it cites, and which pass produced it — text, geometry, or vision. No black box. You can read the rule, see what the engine looked at, and decide whether it's a real catch.
There is a strong temptation in our industry to frame AI tools as a way to reduce billable engineer hours. We refuse that frame. Engineering hours are how our industry creates value, and reducing them is reducing the firm's revenue while transferring professional responsibility to a piece of software that cannot legally hold it.
Stratum Check does not design. It does not stamp. It does not sign off. Every finding it produces goes to a licensed engineer for review, confirmation, or dismissal. Your engineer reads every sheet, makes every call, and signs every submittal — exactly as they did before.
What changes is what their attention goes to. The mechanical checklist pass — every published rule on every sheet — happens in advance. By the time the engineer opens the package, the systematic findings are already surfaced and triaged. Their hours go to the design trade-offs, the ambiguous calls, and the judgment work that actually requires their license. Junior staff get the citation-backed feedback loop that builds real fluency in the standards. The whole firm gets better.
There is a strong pull right now to treat AI as a universal answer in engineering software — drop the plan set in, ask a model, accept the response. We don't build that way. Stratum Check is meaningfully an AI product, but the AI is scoped, not general.
Rules are engineered, not generated. Every check is hand-built by working California civil transportation engineers from published agency standards, validated before going live, and runs as deterministic code — same input, same output, every time. AI may help engineers draft candidate rules from new standards documents, but the runtime engine never asks a model what the rule should be.
AI is reserved for routing and vision. The model classifies each sheet so the right rules apply, and performs the subset of checks where pixels are the only meaningful signal — drafting symbology, plotted annotation matching, vision-pass verifications a human reviewer would also do with their eyes. The model is asked precise, scoped questions tied to specific rules. Never open-ended judgment.
This is the line we hold. Wherever a check can be expressed as deterministic engineering code, that's how it runs — and AI is the backstop for what genuinely needs vision. Every finding traces back to the rule that produced it, the standard it cites, and the pass that ran it. The system tells you not just what it found, but why and how. That's what an engineer's tool should do.
Stratum Check exists because we've spent fifteen-plus years delivering California public infrastructure ourselves — and watched, again and again, how the systematic QC checks slip when deadlines compress and reviewer attention runs out.The Stratum Check team
Project management, design, and construction across highway, transit, rail, and flood projects throughout the state.
Licensed professional engineers and licensed land surveyors. Project Management Professionals. Certified Floodplain Managers. The credentials we ask of our customers' reviewers.
Personal involvement in projects ranging from complete-streets corridors to $500M+ interchange projects and regional rail extensions.
Running plan sets through Stratum Check produces more than findings. Over time, a feedback loop forms — junior staff learn the standards in context, and senior engineers spend their reviews on the calls that actually require a stamp.
Every finding cites the exact PPM or CADD section it came from, with context explaining when the rule applies and why. Running their own plan sets through Stratum Check, junior staff pick up the standards in real situations — faster and more durable than reading the manual cover-to-cover, and more reliable than catching things ad-hoc under deadline.
The systematic checklist pass happens before your senior reviewer opens the plan set. Their attention stays on ambiguous calls, design trade-offs, and the handful of findings that actually need a licensed decision — not on hunting for the 40th missed postmile annotation of the day. Senior engineer time stays on senior engineer work.
Currently in closed beta. Get in touch to discuss access and pilot terms.
Get started →