AI photometric engine · parking and roadway lighting
Give it the site, a luminaire and your criteria. It searches thousands of pole layouts and returns one that passes, with every input on record so you can check it in AGi32.
Runnax is an AI photometric engine for roadway and parking lighting. Give it a site, a luminaire and your criteria; it returns a pole layout that passes, with every input on record so you can check it in AGi32.
Watch it lay out a 300 × 200 ft lot. Poles may only go on the islands and curbs (teal).
Each ships when it passes the same AGi32 comparison the parking engine did.
Hand it every IES file you would quote. It runs the full search for each one and keeps the luminaire that passes with the fewest poles.
You mark the islands and curbs, or confirm areas picked from your DXF. Any layout with a pole outside them is thrown out: 18,721 of 24,904 in this run.
Pole table, point-by-point grid, isolux sheet and a summary with every input. Put the pole table on your sheet. Every input is in the summary, so you can rebuild the run in AGi32 for a second opinion.
One contract behind a REST API, an MCP server and a CLI, so an AI assistant gets the same numbers an engineer does. Early access runs through us today.
Same lot, same luminaire table the engine used, same point method, running in your browser. Drag a pole along the islands, change the height, watch the grid. This is a preview; the engine's run above is the record.
Luminaire: the one the engine picked for this lot, measured IES. Grid 10 ft, direct light only, same point method as the engine.
Engine grid, 271 points at 20 ft, 15 heads. Against an AGi32 direct-light run of the same lot, min, average and max agree within 0.5 %.
The engine reads 1.7 % higher on average than the delivered design; the tighter of the two leaves 3.04 % headroom on uniformity. Direct light only; not an RP-8 compliance statement; headroom is specific to this lot.
| Manufacturer free-layout service | Runnax | |
|---|---|---|
| Turnaround | 1–3 business days, by a person in AGi32 | about a minute |
| Same 300 × 200 ft lot | 24 × 150 W, no pole positions, 50 % over target | 8 × 166 W, 8 poles on the islands, on target |
| Can you check it? | No inputs provided | Every input on record; checked against AGi32 on a real parking project |
Left: public free-layout tools of two US LED manufacturers, tested 2026-10-08. Their tool optimises average illuminance, ours minimum, so this compares capability, not scores.
The site (a DXF, an AGI file, or the lot dimensions and drive aisle), the luminaires you would quote as IES files, your targets (minimum illuminance and max:min uniformity), a pole-height range, and where poles may stand: the islands and curbs.
Parking: on a real lot, the engine's minimum, average and maximum are within 0.5 % of an AGi32 direct-light run of the same lot over 271 points; against the AGi32 design delivered to that customer, its average reads 1.7 % higher. Roadway: area averages are within 3.6 % of AGi32 and DIALux on 9 zones from real projects. Roadway point values and minimums differ more, so we don't quote them on their own.
No. It checks the targets you set, using direct light only. A run is not an RP-8 compliance statement and not a construction approval; the engineer of record still signs the design.
Only inside the areas you mark as allowed, such as islands and curbs. It searches regular arrays of up to 16 poles in up to 3 rows. Any layout with a pole outside the allowed areas is rejected.
Parking: curved or polyline aisles, mixed mounting heights in one lot, several parking zones in one run, obstructions and shadowing, and aiming optimisation (single-head fixtures keep one fixed orientation for now). Roadway: staggered arrangements, three or more roadway lanes in one section, luminance-based criteria, and pass/fail verdicts. Nothing is field-verified against measured light levels yet; all results are simulations.
No. Files are used for your run, deleted after 30 days, and never used for training.
The engine has one contract behind a REST API, an MCP server and a CLI, all shipping with the API release. Until then, early-access runs go through us.
Runnax is a photometric engine company founded by Kyan Gao. Auvolar, a commercial LED manufacturer in California, is its first customer.
If you'd rather describe the project first, or you're evaluating the API or a white-label deployment, write to the founder directly.
Replies within one business day, US Pacific.