ML
Mannapov, LLC
📋
Description
- We’re looking for a hands-on AI Automation Intern to build a real-world system that uses large language models (LLMs) and n8n to handle sales communications across Twilio (voice/SMS/WhatsApp) and O365 email, with all activity logged in HubSpot. You’ll work directly with the CTO and sales leadership to design, build, and refine workflows where AI can draft and eventually send replies, answer questions about pricing, negotiate simple terms, and escalate more complex situations to human reps. This is not a research-only role—you will ship something that gets used.
- Design & Build n8n Workflows:
- Create and maintain n8n workflows that ingest calls, SMS, WhatsApp messages, and emails via Twilio and O365
- Trigger LLM calls (OpenAI / Gemini) for classification, response generation, and summarization
- Log summaries and outcomes into HubSpot / Timelines.AI
- Integrate and Orchestrate LLMs:
- Use different LLMs for different tasks (e.g., classification vs full responses)
- Experiment with OpenAI and Gemini to balance quality, latency, and cost
- Build reusable prompt templates and system messages
- Prompt Engineering & Guardrails:
- Design prompts for intent detection, replies, and conversation summaries
- Implement “do not” constraints and red lines (e.g., pricing boundaries, prohibited promises)
- Encode escalation logic for tricky or high-value interactions
- Human-in-the-Loop to Automation:
- Start with flows where AI drafts a response and a human approves it
- Gradually introduce auto-send for low-risk scenarios as performance improves
- Log and analyze interactions to understand where the AI struggles or needs improvement
- Evaluation & Metrics:
- Build lightweight evaluation tools and test sets of messages
- Help define and track metrics (AI vs human handled, response time, escalation rates)
- Document & Handoff:
- Document architecture and n8n flows
- Document the prompt library and how to update it
- Prepare setup guides and runbooks
- Present your work and recommendations at the end of the internship
- What You’ll Learn:
- How to build production-adjacent AI systems, not just playground demos
- How to orchestrate multiple channels (voice, SMS, WhatsApp, email) with automation tools
- How to design guardrails and escalation strategies for AI in real sales workflows
- How to balance model quality vs cost vs latency in a real business environment
- How AI integrates with CRMs (HubSpot) and real sales operations
🏖️
Benefits
- Hands-on experience building real-world AI systems
- Exposure to multiple communication channels (voice, SMS, WhatsApp, email)
- Learning how to design guardrails and escalation strategies for AI in sales workflows
- Opportunity to work directly with CTO and sales leadership
- Access to real sales data, communication logs, and existing tools (Twilio, O365, HubSpot)
- Guidance on business rules, pricing ranges, and boundaries