Narendra Singh

Conversational AI and Voice AI Engineer who designs and builds real-time chatbot, voicebot, and voice AI platforms end-to-end, from ground-zero architecture to production deployment. Experienced across the full conversational pipeline – telephony/SIP ingress, WebRTC media layer, streaming STT/TTS orchestration, NLP-driven dialog management, backend APIs, and databases – with hands-on ownership of scalability, latency, and fault tolerance. Comfortable working directly with cross-functional and customer-facing teams to translate requirements into production solutions, debug live bot behavior, and optimize user experience.

Has taken multiple voice/chat AI platforms from a blank repository to production on AWS, and is an active open-source contributor, including to LiveKit.

AI Python Developer - Exalogic Consulting Pvt. Ltd. - Bengaluru, India

(2025-12)

• Architected and built a real-time conversational voice AI interview pipeline from scratch – LiveKit media layer, STT/TTS orchestration, NLP-driven dialog management, and session management – achieving sub (700-800) ms response latency and 95%+ transcription accuracy.
• Designed the end-to-end system architecture for a full-stack recruitment automation platform (API layer, database schema, RAG retrieval layer), reducing recruiter screening time by 60% and processing 10,000+ resumes/month.
• Made core architectural decisions on service boundaries and data flow for resume parsing, duplicate detection, and candidate lifecycle management, then implemented the resulting backend services and APIs.
• Designed candidate-scoring and feedback-generation subsystems and integrated them into the platform architecture.
• Debugged and iterated on the conversational pipeline architecture to optimize speech recognition, improve bot response quality, and cut end-to-end latency in production.

Data Scientist - Bluebash Consulting Pvt. Ltd. - Mohali, India

(2024-05 - 2025-12)

• Designed the end-to-end architecture for a multilingual conversational voice AI platform from scratch – media layer (LiveKit/WebRTC), STT/TTS pipeline, and dialog/conversation state management – then built and deployed it, achieving sub (700-800) ms response latency.
• Architected the telephony ingress layer, building SIP-based VoIP integrations with Twilio, Jambonz, and telecom providers to support seamless inbound and outbound voicebot interactions.
• Designed and built a full-stack conversational voice assistant platform ground-up, defining the system architecture connecting backend services, OpenAI, Pinecone, LiveKit, Twilio, STT, and TTS to deliver context-aware, domain-specific bot responses.
• Revisited and re-architected retrieval, NLP intent-matching, and vector search components of the pipeline, reducing transcription latency by 30% and improving response accuracy and conversational quality.
• Designed backend tooling, workflow automation, and AI-to-human handoff architecture, improving enterprise operations, escalation handling, and business analytics – while acting as point of contact for stakeholders on bot behavior and issue resolution.
• Owned the reliability architecture end-to-end fault-tolerant voice-agent design, fallback mechanisms, monitoring, and scalable deployment topology for enterprise conversational AI applications.

Bachelor of Technology - Data Science and Artificial Intelligence - Indian Institute of Information Technology Dharwad (IIIT Dharwad) (2020 - 2024)