Voice AI Engineer Kollaikal Rupesh joins Antler San Francisco residency

By AI, Created 8:10 AM UTC, May 20, 2026, /AGP/ – Kollaikal Rupesh has been selected for Antler’s 2026 San Francisco residency after work that improved open-source voice infrastructure and cut streaming latency at Wayline. He is now building Miranda AI, a legal voice platform that answers calls, handles intake, and drafts documents live.

Why it matters: - Rupesh’s work targets a core problem in voice AI: systems that sound impressive in demos but still fail in production. - His endpoint detection and latency improvements affect how reliably voice agents handle real phone calls, especially in telephony and legal workflows. - Antler’s selection gives Miranda AI a stronger path to build and commercialize a product aimed at law firms.

What happened: - Kollaikal Rupesh was selected for Antler’s 2026 San Francisco residency. - Rupesh joins Antler SF after serving as Founding AI Engineer at Wayline, a Y Combinator S25 company. - Rupesh is also the founder of Miranda AI, which was built for the Stanford CodeX LLM x Law Hackathon #6 in April 2026. - Miranda AI placed second in the hackathon. - The hackathon drew more than 650 attendees and was judged by representatives from Pear VC, Baker McKenzie and Stanford Law faculty.

The details: - Rupesh contributed to Pipecat, an open-source framework for real-time voice agents with more than 12,000 GitHub stars. - A high-fidelity audio resampling change improved Pipecat endpoint detection accuracy on telephony audio from 59% to 94%. - Additional Pipecat pull requests covered runtime failover behavior, heartbeat stability and serializer reliability. - Rupesh also wrote Pipewatch and Coldcall, two open-source evaluation tools for latency and transcript reliability in streaming voice pipelines. - At Wayline, Rupesh led streaming speech inference work that cut end-to-end conversational latency from 4–5 seconds to under 600 milliseconds. - Rupesh built Wayline’s evaluation framework for word error rate stability, transcript drift and latency at P50 and P95. - Rupesh became Wayline’s top contributor after joining in early 2026. - Before Wayline, Rupesh was a Forward Deployed Engineer at Smallest AI, where he led production voice agent deployments for enterprise customers across multiple industries. - Rupesh’s published work includes a sole-authored survey on turn detection in production voice AI agents. - Rupesh also authored a paper on deployment-first voice clone detection for audio deepfake defense and has peer-reviewed work in AIP Conference Proceedings. - Rupesh has presented at AWS San Francisco on real-time healthcare AI agent design. - Rupesh co-hosted AI Dev Tools Night at Cloudflare HQ with Sourcegraph, drawing 251 attendees. - Rupesh chaired Aggie Hacks 2025 at UC Davis. - Miranda AI now answers calls, runs legal intake with statute-of-limitations math and conflict screening, drafts engagement and demand letters live, and delivers qualified matters to attorneys. - Miranda AI is being deployed across plaintiff, litigation and immigration practices. - Miranda AI positions itself as an AI operating system for plaintiff, litigation and immigration law firms. - The company’s product uses one phone line to answer every call, run intake, draft documents live and route qualified matters to attorneys. - More information is available on Miranda AI’s website.

Between the lines: - Rupesh’s résumé is centered on infrastructure rather than model hype, which matches a broader shift in voice AI toward reliability, measurement and deployment readiness. - The Antler residency suggests investors see legal voice automation as a practical wedge for real-world adoption, not just a research demo. - The combination of open-source contributions, startup execution and published research gives Miranda AI both technical credibility and a product story.

What’s next: - Rupesh is expected to use the Antler residency to build Miranda AI further and sharpen its go-to-market strategy. - Miranda AI’s near-term focus appears to be broader deployment across plaintiff, litigation and immigration firms. - Rupesh is also likely to continue work on voice AI reliability tools and infrastructure as the market moves from experimentation to production.**