Prompt Engineer AI Voice & Conversational Systems
Cloud Builders Technologies · Jun 2025 – Jun 2026
- Designed, tested, and optimized LLM system prompts for real-time voice AI agents handling live customer calls per month across GPT, Claude, and Gemini - benchmarking each model's accuracy, latency, and cost for different conversational use cases.
- Restructured system prompt architecture and slot-filling logic for appointment-booking agents, reducing call drop rate by 20-30% and improving booking accuracy by 85%.
- Built AI receptionist systems integrating Vapi, Twilio (SMS), Google Calendar, and CRM platforms for fully automated inbound/outbound call handling and lead management.
- Mapped conversation flows, fallback strategies, and edge cases for voice bots and IVRs, balancing UX (tone, clarity, accessibility) against model constraints (latency, interruption handling, context limits).
- Engineered multilingual prompt variants for regional-language voice agents, improving comprehension accuracy and customer satisfaction for non-English callers.
- Partnered with engineers to productionize prompts inside voice AI pipelines (Vapi, Twilio, ElevenLabs APIs), maintaining zero-downtime deployment across [X] client environments.
- Built and maintained n8n automation workflows connecting CRM, scheduling, and messaging systems, enabling reusable, client-agnostic deployment templates.
- Diagnosed failed/dropped calls by isolating root cause across the pipeline (STT transcription errors, LLM reasoning, tool/API calls, latency), reducing misattributed 'prompt bugs' and speeding up fix turnaround.
- Used Postman to test and debug REST API integrations (Vapi, Twilio, CRM, Calendar) - configuring request parameters, headers, and auth, and validating responses before wiring endpoints into production voice/automation pipelines.
- Built and maintained a prompt versioning process (changelog, rollback points, regression test suite of standard conversations) so every prompt update was benchmarked against booking accuracy, drop rate, and hallucination rate before full rollout.
- Authored and maintained a prompt library and testing documentation (conversation specs, QA results, design rationale) used across the team for consistency and faster onboarding.
- Owned client onboarding and go-live setup for AI agent deployments, improving retention through proactive support and iterative prompt tuning post-launch.