Bidflow Uses Computer Vision to Count Every Socket on a Floor Plan

The Macro: Electrical Estimating Is Still Done by Hand With a Highlighter

Construction bidding is a numbers game. Electrical contractors look at architectural drawings, count every outlet, switch, junction box, and light fixture, estimate labor hours, price out materials, and submit a bid. The company that does this fastest and most accurately wins the job.

The counting part is called a takeoff. It is mind-numbingly tedious. An estimator opens a PDF floor plan, zooms in, counts symbols, marks them with a digital highlighter, records quantities in a spreadsheet, and repeats for every page of a drawing set. A large commercial project can have hundreds of pages. The process takes days.

This is exactly the kind of work where computer vision should excel. Electrical symbols on floor plans are standardized. The variations are manageable. The output is a count of known objects in known categories. It is object detection applied to a structured domain with clear ground truth.

Yet most electrical contractors still do it by hand. The existing estimating software like Accubid, ConEst, and PlanSwift handles pricing and proposal generation. But the actual takeoff, the counting, is still largely manual. Some tools have basic measurement capabilities, but none have cracked AI-powered symbol detection at production quality.

The Micro: A Teenage Coder and a Teenage Entrepreneur

Jesse Choe and Gautham Ramachandran founded Bidflow. Jesse dropped out of college, ranks in the top 1% of competitive coders in the US, and was a Jane Street extern. Gautham bootstrapped a company to $120K revenue by age 16. They are a two-person team from New York, part of YC Winter 2026 with Ankit Gupta.

The product detects electrical symbols on floor plans with 95-99% accuracy for power and lighting takeoffs. It includes PDF tools for drawing navigation, measurement, and correction. Detected quantities automatically convert to cost estimates. A premium tier at $50/month adds an AI assistant that answers questions about drawings and specifications, plus automated data entry for material pricing and labor units.

The pricing is aggressive and smart. Pay-as-you-go at $0.03 per correctly detected symbol. They claim this is 70% cheaper than manual counting at roughly $0.10 per symbol, and 20x faster. For an electrical contractor deciding whether to try the tool, the risk is essentially zero. Pay three cents per symbol. If the detection is wrong, you do not pay.

The turnaround is fast. The go-to-market targets both lighting distributors who need to quote quickly and electrical contractors who want to bid more jobs.

The Verdict

Bidflow is a clean vertical AI play. Specific problem, measurable value, clear buyer. The per-symbol pricing model is brilliant because it aligns cost with value and removes adoption risk. If the detection accuracy holds at 95%+, this product sells itself.

The risk is that the big construction software players add similar features. Procore, Autodesk, and Trimble all have the resources and customer relationships. But construction software moves slowly, and these incumbents are focused on broader platform plays rather than vertical-specific AI.

In 30 days, I want to see the accuracy rate across different drawing styles and qualities. Blueprint quality varies enormously. In 60 days, the question is customer retention. Are contractors coming back for every bid or just trying it once? In 90 days, I want to know about expansion beyond electrical. If the computer vision works for electrical symbols, it can work for plumbing, HVAC, and fire protection. Each trade is a new market with the same product architecture.