Olek Paraska

AI Summary

→ WHAT IT COVERS Olek Paraska, CTO of Togal AI, explains how computer vision and AI agents automate construction estimating, a manual process that delays projects by years. Togal tripled revenue three consecutive years by solving the takeoff bottleneck where contractors manually measure floor plans. The company targets preconstruction workflows to accelerate building timelines and reduce costs. → KEY INSIGHTS - **Manual Takeoff Elimination:** Construction estimators currently spend full days manually tracing room dimensions and counting doors on floor plans for every project. Togal's proprietary computer vision models extract measurements automatically, reducing this process from one day to under one hour, achieving 90% time savings. Every subcontractor previously measured the same spaces independently, creating massive redundancy across hundreds of contractors per building. - **Training Data Challenge:** Building accurate construction AI requires annotated floor plans from professional architects, not general crowdsourced workers, because construction drawings contain specialized nuances. Togal annotated thousands of real floor plans and developed synthetic data generation from CAD software. However, real-world messy construction data consistently outperforms synthetic data, as machine learning models can distinguish between them and synthetic data produces inferior results. - **Agentic Preconstruction Workflows:** AI agents now handle routine tasks like generating RFIs (requests for information), comparing floor plan versions to identify meaningful changes, and parsing 500-page specification documents to extract relevant scope. Agents connect perception layers (reading floor plans) with reasoning layers (large language models) to answer questions like what materials are missing or how to reduce costs while maintaining specifications. - **Construction-Software Culture Gap:** Togal employs both construction professionals working in their first software company and software engineers in their first construction company, creating internal communication challenges that mirror industry-wide technology adoption barriers. Construction resists bad technology, not technology itself, because physical world consequences make the barrier to entry extremely high. Solutions require deep construction domain expertise combined with technical capability. - **Revenue Growth Indicators:** Togal's annual revenue tripling for three consecutive years reflects industry hunger for solutions rather than typical startup growth from a small base. The company focuses on commercial buildings like hotels and hospitals, including most Miami high-rises. Construction productivity has stagnated or declined over fifty years while other industries advanced, creating massive opportunity for AI-driven efficiency gains. → NOTABLE MOMENT Paraska reveals that some construction estimators still print architectural drawings and measure rooms with physical rulers on paper rather than using digital tools. This analog approach persists in an industry building multimillion-dollar structures, illustrating how construction operates decades behind software industries in adopting basic digital workflows, let alone advanced AI capabilities. 💼 SPONSORS None detected 🏷️ Construction AI, Computer Vision, Preconstruction Automation, Building Estimating, Agentic Workflows