I own a large dental practice and currently spend significant time each week manually comparing supply prices across publicly available competitor websites to negotiate price matches with my primary supplier (Henry Schein Dental). I need an automated agent to do this for me. THE WORKFLOW: Each week I generate an order PDF from Henry Schein’s online ordering system. The agent should: 1. Parse the PDF to extract a structured list of items (product number, description, quantity, unit price) 2. Run a broad market price sweep using publicly available shopping and search tools to establish a price floor for each item 3. Search a defined and extensible list of publicly accessible dental supplier websites for each item 4. Return a clean comparison spreadsheet showing the best publicly available price found per item, sorted by savings opportunity This is Stage 1 of a planned 4-stage system. A full project specification document will be provided to candidates under serious consideration. IMPORTANT — PUBLIC PRICING ONLY: This system works exclusively with publicly posted pricing from websites that display prices without requiring a login or account. No authentication or credential handling of any kind is involved. If a site requires login to view prices, the agent flags it and skips it automatically. All data accessed is publicly available to any web visitor. WHAT MAKES THIS PROJECT DISTINCTIVE: - A broad market price sweep runs first for each item to establish a baseline before hitting individual supplier sites - Search logic must distinguish between exact product matches and quantity-contingent pricing, with different handling for each in the output - The site list is maintained in a plain-text configuration file — the client must be able to add new supplier sites without touching any code - The system needs an equivalency framework: some products require exact brand and SKU matches; others can be compared against generic alternatives. A client-maintained reference table drives this logic - Sites that require login are flagged and skipped automatically — no access is attempted DENTAL SUPPLIER SITES TO SEARCH (all publicly priced — no login required): Net32, Crazy Dental, Anson Dental Supply, MVP Dental, Dental Market, Pricenex, TDSC, Safco Dental, Spark Dental, Dental City, Optimus Dental Supply, Medex Supply, My DDS Supply, Pairey Dental, Artcraft Dental, AZDental.com, AliExpress, and others via the extensible configuration file. Additionally, the agent uses publicly available shopping and search tools to establish a market price floor for each item before searching the supplier list. OUTPUT: Excel spreadsheet — best publicly available price per item, source, savings per unit, total savings. Sorted by savings descending. A separate background file captures URLs and quantity-condition detail not included in the primary report. TECH STACK: I’m open to your recommendation. Many supplier sites use JavaScript-rendered pricing, so the solution must handle fully rendered pages. Claude API (Anthropic) is preferred as the AI reasoning engine for product matching and description parsing. Local execution on macOS. IMPORTANT — ARCHITECTURE AND FUTURE STAGES: This is Stage 1 of a deliberately phased 4-stage system. Stage 1 is the only contracted deliverable, but the developer must understand and design for the full roadmap so that foundational decisions made now do not require a rebuild later. The three future stages are fully described in the specification document. In summary: • Stage 2 adds full equivalency table integration and a second output stream — a direct purchase list for items that should be bought from alternative suppliers rather than price-matched with the primary supplier • Stage 3 adds order generation assistance — the agent imports Henry Schein order history and helps draft the weekly order list based on usage patterns and par levels, feeding into the Stage 1 workflow • Stage 4 adds inventory intelligence and reorder forecasting — frequency analysis, consumption rate modeling, reorder projection, and quantity optimization Bids that do not acknowledge the phased architecture and explain how Stage 1 will be designed to accommodate it will not be considered. The goal is to build once correctly, not rebuild repeatedly. YOUR PROPOSAL SHOULD INCLUDE: - Proposed tech stack - How you will handle JavaScript-rendered pricing pages - How you will handle sites where public pricing is unavailable - Milestone breakdown and timeline for Stage 1 - Examples of prior browser automation and AI agent projects Two sample Henry Schein order PDFs (one clean, one annotated with my current manual process) will be provided to the selected developer.