Tasks
85 tasks across 4 engineering disciplines (29 in the public dev set, 56 in the hidden test set), with a 6-hour time limit per task.
Structural engineering
10 tasks, 5 graded by automated checks and 5 by both checks and LLM-Judge.
Calculate gravity, wind, and seismic loads, verify structural members and connections, and design roof decks, diaphragms, and equipment supports.
Illustration: Roof members under load.
Public dev tasks (4)
| Task | Graded by |
|---|---|
| Engineer-of-record corrections Incorporate engineer-of-record markups into an issued drawing revision across two bid packages, including comment dispositions, sheet index updates, and change documentation. |
Automated checks and LLM-Judge |
| Wind diaphragm shear Compute the wind roof-diaphragm shears and shear-wall loads for a single-story data hall. |
Automated checks |
| Seismic diaphragm shear Compute the seismic diaphragm shears and shear-wall loads for a data-center lateral force-resisting system. |
Automated checks |
| Roof deck design Design a data center's steel roof deck for gravity and lateral loads, specifying gauge, fastener pattern, and diaphragm capacity. |
Automated checks |
6 more tasks in the hidden test set.
Civil engineering
18 tasks, 5 graded by automated checks, 10 by LLM-Judge and 3 by both.
Lay out site grading, drainage, and utilities, evaluate soil settlement and foundation strategies, and verify excavation shoring and construction plans.
Illustration: Buildings and utility routes within site limits.
Public dev tasks (9)
| Task | Graded by |
|---|---|
| Phase-3 landscape revision Incorporate a street-tree review comment into a landscape drawing set and check the package for agency resubmittal. |
LLM-Judge |
| Data-center site design Generate an IFC site model for a two-phase data-center campus that satisfies all requirements in its approval document. |
Automated checks |
| Phase-3 engineer's estimate Verify a project section of an engineer's estimate against the construction plans before sending it to the surety broker. |
LLM-Judge |
| Block C footing settlement Verify the long-term settlement of a building block's spread footings on clay against the project record and issue the verification memo. |
LLM-Judge |
| Frame extension method package Plan a steel frame extension over an occupied floor and check the construction method package against applicable safety regulations. |
Automated checks |
| Foundation strategy Select spread footings or aggregate piers for six building pads from geotechnical records and deliver a design basis, schedule, and constraint register. |
Automated checks and LLM-Judge |
| Fill compaction acceptance Check slab-on-grade fill compaction test records against the earthwork specification and issue a materials acceptance record. |
LLM-Judge |
| Block B shoring verification Verify a revised excavation shoring design against logged groundwater and subsurface conditions prior to excavation. |
LLM-Judge |
| C-9 emergency spillway check Evaluate a stormwater basin's emergency spillway drawing against a plan-check comment and specify the required revisions in a memo. |
LLM-Judge |
9 more tasks in the hidden test set.
Electrical engineering
27 tasks, 11 graded by automated checks, 14 by LLM-Judge and 2 by both.
Analyze power distribution from utility service to branch circuits, size feeders and cables, review solar interconnections, and check drawings for energization.
Illustration: A switchboard and its connected loads.
Public dev tasks (9)
| Task | Graded by |
|---|---|
| Supplemental interconnection review Write the supplemental interconnection review memo for a solar application on a distribution feeder. |
LLM-Judge |
| Municipal witness-test package Prepare the witness-test scope and schedule for municipal solar arrays that require testing, and identify those that do not. |
Automated checks and LLM-Judge |
| Conveyance study Complete the conveyance study and cost workbook for seven medium-voltage circuits in a 13.8 kV site distribution system. |
Automated checks |
| Device takeoff Complete an electrical device takeoff from a data hall's drawing set and price the bill of materials. |
Automated checks |
| Electrical specifications Write the Division 26 electrical specifications for an Army facility from its coordinated design package. |
Automated checks |
| ALN-108 pre-application response Draft a utility response to a developer's pre-application capacity request for a distribution feeder and close out the tracking record. |
LLM-Judge |
| Corwyn permission to operate Determine whether a utility operations center can grant permission to operate for a PV site on the requested date and initiate the work package. |
LLM-Judge |
| Halbray backfeed reply Assess whether a community solar site can be energized for backfeed on the requested date, draft the developer response, and initiate the work package. |
LLM-Judge |
| Low-voltage cable sizing Size a building's low-voltage wire schedule for voltage drop, duct-bank ampacity, and cost, complete with a constraint register and design basis. |
Automated checks and LLM-Judge |
18 more tasks in the hidden test set.
Mechanical engineering
30 tasks, 8 graded by automated checks, 18 by LLM-Judge and 4 by both.
Design and troubleshoot HVAC, thermal, and fluid-flow systems, route clash-free building services, and validate equipment performance and fire simulations.
Illustration: Heat removal within a cooling loop's flow limit.
Public dev tasks (7)
| Task | Graded by |
|---|---|
| Fire-test validation memo Evaluate whether a fire-resistance simulation is validated against furnace test data and write an engineering validation memo. |
LLM-Judge |
| MEP coordination, hall A Reconstruct a data hall from seven drawing sheets and generate a clash-free IFC4 model routing cable tray, busway, and chilled-water piping. |
Automated checks |
| MEP coordination, hall B Route MEP systems through a data hall where the structural grid is offset from the floor tiles, clearing a cross-aisle and 104 sprinkler heads. |
Automated checks |
| Non-converging fire-solve disposition Issue a signed disposition on a fire simulation's non-converging steady solve before a fourth run begins. |
LLM-Judge |
| LES mesh-resolution note Document the mesh-resolution basis of a large-eddy fire simulation to close out a peer-review comment. |
LLM-Judge |
| Cooling-tower life-cycle cost Design a condenser-water heat rejection replacement for a federal campus and justify the selection with a life-cycle cost analysis. |
Automated checks |
| PCW flush procedure Develop the cleaning, flushing, and passivation procedure for a data hall's process cooling water loop. |
Automated checks |
23 more tasks in the hidden test set.