Member of Technical Staff, Cekura (Bangalore, In-Person)

• Simulation engine. Simulate thousands of realistic conversations across voice, chat, and phone, with control over personas, interruptions, noise, and edge cases.
• Agent evaluation. Build LLM evaluators, metrics, red-teaming, drift detection, and the autonomous loop that detects, reproduces, fixes, tests, and ships a PR.
• Speech-to-speech research. Work with labs to find where models fall short of human conversation, then build the benchmarks, datasets, reward signals, and RL environments that close the gap.
• Applied audio. Turn the paralinguistic layer (emotion, tone, silences, hesitations, overlaps) into structured signals, so models are judged on how things are said, not just what.
• Real-time voice infra. SIP, WebRTC, STT/TTS, Twilio, Vapi, Retell, LiveKit, Pipecat. Latency and barge-in are first-class problems.
• Ship end-to-end and shape the team. Own what you build from design to production, and set the standards a growing team inherits.

• A strong generalist who wants to work across systems and AI, not stay in one lane.
• A strong coder in Python, TypeScript, or Go, with experience in distributed systems, real-time infra, LLM products, or model training and evaluation.
• You've run production systems and care about reliability, latency, and correctness.
• Sharp instincts about where LLMs and speech models fail, and how to build reliable systems on them.
• You like ambiguity, move fast, communicate clearly, and raise the bar around you.
• You read the first section twice and you're still here.
• Bonus: ex-founder or aspiring one.

• LLM evals, agent frameworks, or AI observability.
• Voice AI stack experience (Twilio, SIP, WebRTC, Vapi, Retell, LiveKit, Pipecat, STT/TTS).
• RL for language or speech models, benchmark building, or speech-to-speech models.
• Early-stage startup experience or open-source work.

We work with unusual intensity: in-person in Bangalore, six days a week, long days, most weekends. That's how we've chosen to build, because in our market speed decides who wins. We say this up front because we want people who feel pulled in, not talked into it. If you want a 9-to-5 (no judgment), this isn't your role.

• You want a narrow role, a fixed stack, or to work only on models or only on infra.
• You need rigid process or heavy structure.
• You're optimizing for work-life balance right now.
• You don't want to work in-person in Bangalore.

• The hardest problems in voice AI, worked on alongside the labs building the models.
• A 12-person engineering team post-seed: early enough to matter, funded enough to move fast.
• Direct work with founders and a highly technical team.
• Top-of-market pay, meaningful equity, all meals at the office, and relocation support.

• Top-of-market cash. Extraordinary hours deserve more than an average salary.
• Meaningful equity. Significant grants, employee-friendly terms, and liquidity opportunities as we raise.
• No commute tax. Support to live near the office, plus dinner there every night.
• Founders in the trenches. We work the same schedule we ask of you.
• A decade of growth in two years. You'll ship more and own more than anywhere else.

Cekura (YC F24) builds evals and improvement for voice AI, covering both voice agents and the speech-to-speech models behind them.

• Voice agents. Teams test before launch, monitor production calls, and improve continuously. When something fails, Cekura reproduces it in simulation, fixes it, tests the fix, and raises a PR. The platform covers simulation, LLM evaluation, red-teaming, drift monitoring, and benchmarking across Vapi, Retell, Pipecat, LiveKit, ElevenLabs and more.
• Speech-to-speech models. We give labs benchmarks, data, and RL environments to improve turn-taking, interruptions, naturalness, and tool use in live conversations.

Customers include Five9, HighLevel, Twin Health, PwC, Deloitte, Jobber, and Jotform. HIPAA, SOC 2, and GDPR are defaults. We're growing fast and backed by top investors.