About the Role

We’re building low-latency, high-reliability voice agents on top of the Lyzr platform.

This role owns the core real-time systems that make live voice conversations feel natural, fast, and dependable. You’ll design the architecture that sits between audio streaming, speech models, and LLMs—where every millisecond matters.

If you care about latency budgets, streaming correctness, and building systems that stay stable under real load, we want to hear from you.

What You’ll Do

Architect and build the end-to-end real-time voice pipeline
Drive latency reduction across audio capture, streaming, inference, and response
Optimize LLM and speech model inference for real-time use cases
Work with research and product on model selection, training, and fine-tuning
Design guardrails for safety, reliability, and failure handling
Set best practices for streaming service design
Mentor engineers and contribute to technical roadmaps

What You Need

3+ years building production distributed systems, ideally real-time or low-latency
Strong proficiency in Python and Go or Rust
Hands-on experience with real-time streaming:

• WebRTC, RTP / SRTP

• Opus

• gRPC streaming

Experience working with speech models and LLMs
Strong performance engineering skills:

• Profiling

• Async I/O

• Batching

• Cache design

Proven track record of shipping reliable, production-grade services

Why Lyzr

Work on real-time AI systems, not offline demos
Own a core piece of infrastructure that defines product quality
Build voice agents meant to operate at production scale
High ownership, deep technical problems, fast-moving team