Claude API / AI Agent Developer - Improve Existing SEO Content Agent + Build 2 Variations

We are looking for an experienced Claude API / AI Automation developer to improve an existing AI content-generation agent and then build two simpler agent variations based on the same architecture. This is not a from-scratch project. We already have an existing Claude-based content agent that is approximately 50% complete, along with examples and a detailed list of required changes. The goal is to audit what is already working, fix the weak parts, and turn it into a reliable end-to-end content generation workflow. The system is used for SEO/affiliate content, so we care particularly about factual accuracy, hallucination prevention, source/evidence handling, natural writing, SEO/AIO quality, consistent formatting, automated QA, and API cost efficiency. Phase 1 - Existing Agent Audit & Optimization - $200 (Until Monday) You will receive the current agent, existing prompts/workflow, examples of expected output, and a detailed change list. Your responsibility is to: - Audit the existing implementation before making major changes. - Identify what is already working and preserve it where possible. - Fix broken or weak prompts, logic, orchestration, and output handling. - Improve factual reliability and prevent unsupported or fabricated claims. - Improve research/source handling where necessary. - Separate content generation from factual/quality validation where appropriate. - Improve output consistency, SEO/AIO structure, and natural writing quality. - Use deterministic validation for requirements that should not depend entirely on an LLM. - Avoid uncontrolled regeneration/retry loops; failed claims or sections should preferably be repaired selectively. - Keep Claude API/token costs reasonable and avoid using expensive models where they are unnecessary. - Deliver a working, improved version of the existing agent and demonstrate the resulting output. Budget: $200 fixed price The objective is not to replace the current system with your preferred architecture. We want someone who can understand an existing implementation, identify the actual problems, and make the minimum necessary changes to improve it reliably. Phase 2 - Two Simple Agent Variations - $100 Each After Phase 1 is successfully completed, we need two much simpler content agents, approximately 800–1,200 words per output. These should reuse the useful components and engineering patterns established in Phase 1 rather than duplicating the entire system. Each agent will have its own content template, intent, formatting requirements, and prompt/logic adjustments, but they are considerably simpler than the Phase 1 agent. Agent #2: $100 fixed price Agent #3: $100 fixed price Technical Experience We Are Looking For Strong experience with Claude API / Anthropic, structured outputs, prompt/system design, Python or Node.js, API integrations, and AI workflow/agent development. Experience with SEO content automation, SERP/source research, fact-checking pipelines, n8n/Make, DataForSEO or similar APIs, programmatic SEO, AIO, and CMS integrations is highly relevant. We are particularly interested in developers who understand the difference between: LLM prompting and building a reliable production AI pipeline. To Apply Please briefly answer: 1. Have you previously improved an existing Claude/LLM agent rather than building one from scratch? Give one relevant example. 2. How would you prevent a content agent from turning third-party information into unsupported facts or fabricated first-hand experience? 3. If factual QA finds only two unsupported claims in an otherwise good article, how would you repair them without regenerating the entire article or creating an uncontrolled retry loop? 4. What would you inspect first when taking over an existing agent that is already partially working? 5. Which Claude models/tools do you normally use, and how do you control API/token costs? Please keep your answers technical and specific. Generic AI-generated proposals will not be considered. Total initial scope: $400 fixed price Phase 1: $200 Phase 2 / Agent #2: $100 Phase 2 / Agent #3: $100 There is potential for additional paid milestones and ongoing work if the first three agents are completed successfully.