AI Multi-Agent Engine Development
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted2 hours ago
We need an AI engineer to build a standalone, modular multi-agent engine with a simple REST API (FastAPI) that we can plug into our existing backend and future apps. The goal is to run parallel research, self-verify findings, synthesize cited reports, and support outreach with capped follow-ups.
Scope of work
- Build a standalone, modular multi-agent engine in Python using LangGraph or CrewAI for stateful orchestration and loop controls.
- Implement parallel research: break user requests into sub-tasks, query search APIs (Serper/SerpAPI), scrape web data, and store findings.
- Implement a self-verification loop: review gathered data for gaps/contradictions/unsupported claims and automatically run capped follow-up research passes when info is missing.
- Generate structured, cited markdown reports by synthesizing verified data.
- Build outreach + capped follow-up: draft personalized outreach messages (Email, WhatsApp, SMS) and handle scheduled, capped follow-ups for unresponsive contacts.
- Expose clean FastAPI endpoints (or webhooks) to trigger jobs, fetch results, and sync data with our backend.
- Add guardrails (hard token budget caps, loop iteration limits, prompt-injection sanitization) and deliver a clean modular codebase with .env configuration plus Swagger/OpenAPI docs.
Additional information
1-Have you built multi-agent systems with feedback loops using LangGraph/CrewAI? Please share relevant repos or examples.
2-How do you implement loop caps and token cost limits to prevent runaway API spend?
3-What is your fixed price and delivery timeline for this standalone module?
Scope of work
- Build a standalone, modular multi-agent engine in Python using LangGraph or CrewAI for stateful orchestration and loop controls.
- Implement parallel research: break user requests into sub-tasks, query search APIs (Serper/SerpAPI), scrape web data, and store findings.
- Implement a self-verification loop: review gathered data for gaps/contradictions/unsupported claims and automatically run capped follow-up research passes when info is missing.
- Generate structured, cited markdown reports by synthesizing verified data.
- Build outreach + capped follow-up: draft personalized outreach messages (Email, WhatsApp, SMS) and handle scheduled, capped follow-ups for unresponsive contacts.
- Expose clean FastAPI endpoints (or webhooks) to trigger jobs, fetch results, and sync data with our backend.
- Add guardrails (hard token budget caps, loop iteration limits, prompt-injection sanitization) and deliver a clean modular codebase with .env configuration plus Swagger/OpenAPI docs.
Additional information
1-Have you built multi-agent systems with feedback loops using LangGraph/CrewAI? Please share relevant repos or examples.
2-How do you implement loop caps and token cost limits to prevent runaway API spend?
3-What is your fixed price and delivery timeline for this standalone module?
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.