About this role
Senior Software Engineer, AI Practice by Numbers (PBN) | Gurugram, India Scope: AI / Conversational Products & Backend Services About Practice by Numbers Practice by Numbers is a dental practice management SaaS platform serving over 1,500 practices across North America — practice management software, VOIP, payment processing, and analytics. We're now expanding into AI-powered automation, building conversational AI products that change how practices interact with their patients. About the Role We’re looking for a Senior Software Engineer with strong, hands-on experience in both backend product engineering and modern AI systems. The ideal candidate can design and build reliable, production-grade services while also developing, integrating, and deploying AI-powered capabilities. As this is a single opening on a small team, we need someone who can take end-to-end ownership across both areas and contribute independently throughout the product development lifecycle. You'll work on our AI Receptionist — a multi-channel conversational AI (voice, SMS, web chat) for dental practices — and the backend services, APIs, and integrations behind it. Hands-on IC role: you own features end to end, including their production behaviour. Real patient-facing traffic under HIPAA constraints, where correctness and latency both matter. Reports to: Lead Engineer — AI Location: Gurugram, India — this role is open in Gurugram only Work Mode: In-office Working Hours: Primarily IST (10 AM – 5 PM), with some evening overlap with US teams (until 9–11 PM IST) as needed What You'll Do Backend Build and maintain backend services and RESTful APIs in Python (FastAPI / Django) Design schemas and write efficient PostgreSQL queries; use Redis for caching and session state Work with async and event-driven patterns — queues, webhooks, WebSockets, background workers Own the operational side: logging, metrics, alerting, debugging production issues Write unit and integration tests for the business logic you ship AI & LLM Own the conversation layer for real-time voice — turn-taking, barge-in, and recovery when the agent talks over a caller or stalls mid-turn Debug non-deterministic failures in production: fabricated confirmations, invented practice policies, dropped context across turns Own patient identity resolution over the phone — misheard and misspelled names, shared phone numbers, multi-algorithm phonetic matching Version prompts and tool schemas like code: canary changes, measure impact, roll back regressions Own the eval and regression suite — golden conversation sets, replay of production calls, CI gating on prompt and model changes Design guardrails for patient-facing interactions: no medical advice, no unverified data disclosure, tool-call authorization, PHI excluded from logs and traces, escalation to a human when the agent should stop Own the model and voice vendor abstraction so a provider change is configuration, not a rewrite Keep voice, SMS, and web chat consistent — shared state and knowledge base, channel- appropriate behaviour Integrations & Data Integrate with practice management systems (Dentrix, Open Dental, Eaglesoft) and internal PBN APIs Implement secure auth flows, including OTP-based patient verification Define SLOs for conversation success and turn latency; own incident response and postmortems for this surface Collaboration Work with Product Management to turn requirements into working software, surfacing edge cases early Participate in sprint planning, standups, code reviews, and product reviews Document what you build; collaborate across time zones with US-based stakeholders Required Qualifications Experience 4+ years of professional software development experience Hands-on experience building backend services and APIs that ran in production Practical experience with LLM-based applications (GPT-4/4o, Claude, or similar) — prompt design, tool calling, handling model output in real systems. Substantial personal or open-source work counts; tutorial-level does not. Experience debugging and improving a system after it shipped Technical Skills Strong Python — our primary language across AI and backend APIs: RESTful services, webhooks, third-party integrations; FastAPI or Django preferred Databases: PostgreSQL — schema design, indexing, query performance; Redis or similar Async Python (asyncio) and event-driven architectures Cloud: working knowledge of AWS (or GCP/Azure) — compute, storage, managed DBs, queues Version control, code review, and CI/CD as normal parts of your workflow AI Domain Understanding Shipped an LLM system to real users and debugged it in production; can describe a failure mode you found and fixed Specific view of what current models are unreliable at, and how that shapes what runs unsupervised versus behind a tool call or human review Multi-turn conversation design — state that survives interruptions, topic switches, and mid-call corrections Experience measuring LLM systems: offline evals, judge model calibration, or production quality metrics Working knowledge of token cost and latency trade-offs, including where streaming helps and where it adds complexity Soft Skills Frame the problem, not just solve it; write down architectural decisions and the tradeoffs you rejected Clear communication with technical and non-technical stakeholders Able to drive your own work to completion without close supervision Comfortable with a fast pace and evolving requirements Willing to work in-office in Gurugram and overlap with US hours when needed Preferred Experience & Skills Conversational AI — chatbots, voice assistants, or IVR Voice/telephony (Twilio, Vonage) or STT/TTS APIs (Deepgram, ElevenLabs, AssemblyAI) LLM orchestration frameworks (LangChain, LlamaIndex) — or a considered view on skippingthem Healthcare / HIPAA compliance knowledge SaaS or B2B product company background; multi-tenant architecture