Job Title

Location:Noida (Onsite) |Experience:3-5 Years | The MissionWe are not looking for a Data Scientist who trains models in a notebook. We are looking for aBackend Engineerwho builds real-time, bi-directional Voice AI agents. You will build the "Brain and Ears" of our fintech voice bots, ensuring they can listen, think, and speak inunder 800ms. What You Will Actually Build (Responsibilities)The Pipeline:Architect full-duplex voice pipelines usingWebSockets(not just REST APIs). ConnectTwilio/Exotelstreams toDeepgram/WhisperandLLMsin real-time. The Speed:Optimize Python code usingasyncioto handle concurrent calls. If the bot lags, you fix it. The Logic:Move beyond simple "Chatbots." Build complex state machines (e.g.,If user says 'Wait', stop talking immediately) using tools like LangChain or custom Python logic. The Stack:Integrate best-in-class Voice APIs (Deepgram, Google TTS, OpenAI Realtime API) and deploy them on AWS/GCP usingDocker. ✅ The "Must-Have" Tech StackIf you don't have these, please do not apply. Python Frameworks:Mastery ofFastAPIorStarlette. (Experience with only Django/Flask isnotsufficient). Real-Time Protocols:Deep understanding ofWebSockets (wss://)and streaming audio data (bytes handling). Concurrency:Proven experience withAsync/Await (asyncio)in Python. You must know how to handle non-blocking I/O. Voice Tech:Hands-on experience withASR(Whisper, Deepgram, Google) andTTS(ElevenLabs, Azure, Polly). LLM Integration:Experience building products withOpenAI API / Anthropic(Function Calling, System Prompts, JSON Mode). Bonus Points (Good to Have)Experience withSIP Trunking(Twilio/Plivo/Exotel). Knowledge ofVector Databases(Pinecone/Qdrant) for RAG. Experience fine-tuning small models (Llama-3, Mistral) for specific tasks. Familiarity withDocker&Kubernetesfor deployment. How to ApplySend your resume +A GitHub linkto a project where you used WebSockets or LLMs. (Resumes without code samples will be prioritized last).*For direct apply send your resume at singh.jyoti@/ +91 8262004185*