Voice AI Engineer - AI Agents / Telephony

Medicount; Chennai; On-site; INR 1,200,000 - 2,400,000; Full time

Benefits offered by this job

Open-source contributions (nice-to-see

GitHub projects

Job summary

Medicount is hiring for an AI voice agent developer in Chennai. You will build and maintain an automated calling agent that interfaces with phone networks, dials insurers, navigates IVR menus, and logs results to a database.

You will implement an AI checker to audit transcripts and escalate doubtful calls to humans. The role requires strong Python, exposure to Node.js/TypeScript, experience with Twilio or SIP, and familiarity with LLM APIs.

Qualifications

• Proficient in Python; Node.js/TypeScript is a plus.
• Proven project with AI voice/chat agent (demo/video) with code.
• Experience with phone tech: Twilio, DTMF, call recording.
• Familiar with a voice-agent framework (Pipecat, LiveKit, Retell, Vapi) or similar.
• Experience with LLM APIs: function/tool calling, JSON output, prompts.
• Strong written and spoken English.
• Experience with PostgreSQL/Supabase, REST APIs, webhooks.

Responsibilities

• Build the calling agent: connect to phone network, dial insurers, navigate IVR, follow scripts.
• Capture results: save claim status, denial code, amount paid, reference & rep name into database.
• Create AI checker: read transcripts, verify against saved answers, escalate doubtful calls.
• Test and measure: run daily test calls, track speed, success rate, cost per call on dashboard.
• Maintain system: adjust agent for menu changes, evaluate insurers, switch off poor performers.
• Follow HIPAA: use approved vendors, keep patient data off laptops, maintain clean logs.
• Document everything so team can operate the system without you.

Skills

Python

Node.js / TypeScript

AI voice agent experience

Twilio/ SIP

LLM APIs (OpenAI/Anthropic)

REST APIs & webhooks

PostgreSQL / Supabase

English communication

Tools

Twilio Programmable Voice

DTMF tones

LiveKit / Pipecat / Retell / Vapi

PostgreSQL / Supabase

OpenAI / Claude

JSON output handling

Job description

• Build the calling agent: connect to the phone network (Twilio / SIP), dial insurance companies, press the right keys on the phone menu (IVR), wait on hold, and talk to the representative using our scripts.
• Capture the answers: save claim status, denial code, amount paid, reference number and representative name into our Supabase / Postgres database in a fixed format.
• Build the AI checker: a second program that reads each call transcript, checks the saved answers against it, and sends doubtful calls to a human for review.
• Test and measure: build a test setup with a fake insurance phone menu, run test calls every day, and track speed, success rate and cost per call on a dashboard.
• Keep it running: fix the agent when an insurance company changes its phone menu, track which insurance companies work well, and switch off the agent for any that do not.
• Follow US healthcare data rules (HIPAA): use only approved vendors, keep patient data off your laptop, and keep logs clean. We will train you on this.
• Document everything so our team can run the system without you.

Skills we need (must have)

• Strong Python (Node.js / TypeScript is a plus).
• Have built at least one AI voice agent or AI chat agent college project, hackathon, freelance or job — that we can try: a phone number we can call, a live demo, or a video of it working, plus the code on GitHub.
• Phone technology: Twilio Programmable Voice or SIP; sending DTMF key tones; call recording.
• One voice-agent framework: Pipecat, LiveKit, Retell, Vapi, or your own Twilio media-stream code.
• Working with LLM APIs (Anthropic Claude or OpenAI): function / tool calling, JSON output, prompt management.
• Curiosity and self-learning: you follow the voice-AI and LLM space (new models, Pipecat / LiveKit releases, Anthropic and OpenAI updates) and have tried the tools yourself.
• Testing habit: you have measured how fast and how often your agent works, and can tell us the numbers.
• PostgreSQL / Supabase, REST APIs and webhooks.
• Good written and spoken English.

Skills that are a plus (good to have)

• Active GitHub with your own projects; open-source contributions; hackathon wins; a tech blog or YouTube channel.
• Built a Telegram / WhatsApp / Discord bot, a voice assistant, or any LLM app with real users.
• Agents that work through phone menus (IVR) and long hold times — insurance, pharmacy, telecom, collections.
• US medical billing knowledge: claim status, denial codes, Waystar or Availity.
• ElevenLabs, Deepgram or Cartesia speech APIs; Retell or Vapi partner experience.
• Any HIPAA, SOC 2 or data-security experience.