Project Overview
In modern construction, civil engineering, and infrastructure procurement, bidding teams are regularly confronted with massive tender packages spanning hundreds or thousands of pages of technical specifications, legal boilerplate, and bills of quantities. Estimators spend days to weeks manually reading documents just to determine if a project is feasible and to calculate an initial rough order of magnitude price.
Favon transforms this workflow: from tender package upload to an auditable, transparent first-price estimate in minutes. Built and launched in an intensive 48-hour sprint at sprintd Zurich 2026 (Bluelion incubator, Zurich), the platform pairs state-of-the-art document extraction LLMs with parametric cost estimation engines to de-risk bidding and unlock unprecedented response velocity.
Problem & Industry Opportunity
Submitting construction and engineering bids carries enormous operational friction and downside risk:
The Bidding Dilemma: Estimators can only bid on a fraction of published RFPs due to time constraints, leaving significant revenue on the table. When bids are prepared under rushed deadlines, missed scope items or misunderstood penalty clauses cause catastrophic budget overruns. Favon gives teams an immediate, defensible “first price” and scope breakdown before committing deep estimation resources.
Core Features & Platform Capabilities
- Automated Tender Document Ingestion: Ingests complex multi-volume PDF specifications, architectural drawings, and RFP addenda with section-aware chunking and hierarchy preservation.
- Intelligent Scope Extraction: Automatically isolates project deliverables, required materials, labor skill sets, project milestones, and penalty clauses with verifiable page citations.
- Itemized First-Price Estimation: Generates transparent cost breakdowns categorized into labor, materials, equipment, site overheads, and contingency margins.
- AI Risk Scoring & Compliance Checks: Flags aggressive liquidated damage clauses, tight completion windows, and high-volatility material dependencies with confidence ratings.
- Historical Benchmark Alignment: Compares pricing output against historic regional tender data and market unit rates to detect outliers before submission.
Technical Architecture & Cloud Stack
Favon was engineered for enterprise durability, data privacy, and rapid response times:
- Frontend: Modern React application built with TypeScript and modern component architecture, providing real-time data visualisations, interactive breakdown donuts, and price sensitivity sliders.
- Backend & AI Pipeline: Python microservices utilizing document extraction LLM pipelines, prompt-engineered extraction schemas, and deterministic mathematical price calculators.
- Cloud Infrastructure: Containerized and deployed on Azure Container Apps with automatic scaling, HTTPS termination, and session isolation.
- Security & Auditability: Every line item in the first price links directly back to the source clause in the original tender package, guaranteeing full auditability.
Team & Hackathon Details
Team Members: Ashwani Kumar, Daria Horica, Reazul Islam, Tomas Strnadel
Hackathon: sprintd Zurich (9–11 October 2026)
Location: Bluelion Incubator, Zurich, Switzerland
Event Page: app.sprintd.org/events/sprintd-zurich
Live Demo: Favon Web Application (Tender 4215ac9814)
Tech Stack: React, Azure Container Apps, LLM Document Extraction, Python, Docker, Cloud Architecture