About this role
Practice by Numbers is hiring a hands-on engineer to own its AI Receptionist, a patient-facing conversational product spanning voice, SMS and web chat, alongside the production backend services and integrations that support it. The role is based in-office in Gurugram and reports to the Lead Engineer — AI. Work is primarily 10 AM–5 PM IST, with evening overlap with US teams until 9–11 PM IST when needed.
Responsibilities
- Build Python backend services and RESTful APIs; design PostgreSQL schemas and performant queries, use Redis for caching and session state, and implement asynchronous, event-driven workflows. Write unit and integration tests and own logging, metrics, alerting and production debugging.
- Own real-time voice conversations, including turn-taking, interruptions and recovery. Diagnose failures such as fabricated confirmations, invented policies and lost context; resolve patient identity when names are misheard or phone numbers are shared.
- Version prompts and tool schemas, canary and measure changes, and roll back regressions. Maintain conversation evaluations using golden sets, production-call replay and CI gates. Apply patient-facing guardrails, including no medical advice or unverified disclosure, authorized tool calls, PHI-free logs and human escalation.
- Keep channels consistent through shared state and knowledge, maintain model and voice vendor abstractions, and integrate with Dentrix, Open Dental, Eaglesoft and internal APIs. Implement OTP-based verification; define conversation-success and turn-latency SLOs and own incidents and postmortems.
- Collaborate with Product Management and US stakeholders; document architectural decisions and participate in planning, reviews and code reviews.
Required: 4+ years of professional software development, production backend/API experience, and practical LLM application experience, including prompt design, tool calling and handling model output. Candidates must have shipped and debugged an LLM system used by real users; substantial personal or open-source work counts, but tutorial-level work does not. Strong Python, PostgreSQL, asynchronous programming, cloud familiarity, CI/CD, evaluation methods, and understanding of model reliability, token costs and latency are required. Independent ownership and clear communication are expected.
Preferred
conversational AI or IVR; telephony or speech APIs; LangChain or LlamaIndex experience or a reasoned alternative; healthcare/HIPAA knowledge; and SaaS, B2B or multi-tenant product experience.