AI agent software architect and full stack developer

Project OverviewWe are seeking a highly skilled, US-based Lead Developer for a 4-to-6-week project building an AI-driven sales automation layer. The core objective of this system is to automate administrative sales tasks—such as top-of-funnel research, routine quoting, and follow-ups—allowing the sales team to focus on high-value conversations. The system will operate seamlessly within our enterprise CRM environment and rely on strict human-in-the-loop approval processes for all external actions. Core ResponsibilitiesYou will lead the development of specialized AI agents designed to handle specific workflow tasks, including:Lead Generation & Research: Integrating data provider APIs and language models to identify target companies and auto-generate preparatory account briefs. Automated Outreach & Tasking: Building systems to draft personalized sequences, categorize incoming request intent, and automatically sync replies or tasks directly to the CRM. Quote & Order Automation: Developing logic that categorizes incoming requests, drafts standard quotes using historical CRM pricing data, and stages approved orders for processing. Retention Modeling: Creating an agent that analyzes historical order cadences to flag dormant accounts and automatically draft personalized check-ins for the sales team to review. Required Technical StackLocation: Must be located in the United States.CRM Integration: Deep expertise working with enterprise CRM APIs (REST, custom records, integration roles); specific experience with NetSuite is highly preferred. AI/LLM Implementation: Proven track record of building applications using Large Language Models, specifically involving retrieval-augmented generation (RAG) over private, structured datasets like product catalogs and past orders. Architecture: Strong ability to translate captured business rules into testable code and build secure, user-friendly approval queues. Project PhasingTimeline: 4 to 6 weeks.Phase 1: Collaborative scoping and process mapping to finalize agent specifications, identify data fields, and set integration endpoints. Phase 2 & 3: Core build phase in a sandbox environment utilizing historical data replays, followed by parallel testing against current human workflows before any live deployment. Does this capture exactly what you need the developer to focus on, or should we also remove the mention of "parsing inbound order-desk emails" from the quote/order workflow