Backend AI Engineer

This is a founding engineer position, not a title. You'll own one or more of our hardest technical problems, from architecture to production and have real decision-making authority from day one.

The problems you'll work on span three deeply interconnected areas:

LLM and ML Systems:

• Fine-tuning and optimising proprietary Small Language Models for Indian financial conversations.
• Building training pipelines on domain-specific datasets.
• Optimising inference for sub-200ms response latency at scale.
• Building intent classification and end-of-utterance detection models.

Voice and Real-Time Infrastructure:

• Pushing the latency envelope on our real-time voice pipeline, WebSocket audio streaming, STT/LLM/TTS orchestration.
• Working on our custom real-time telephony infrastructure, the layer that connects every AI conversation to the phone network at scale.
• Building resilient systems that handle 10,000+ concurrent AI voice calls.
• Solving hard async concurrency problems in Python at the network and application layer.

Distributed Intelligence:

• Architecting data replication pipelines, scheduling systems, and observability infrastructure that handle millions of conversations daily.
• ClickHouse-backed analytics, Kubernetes autoscaling, and distributed systems thinking that keeps everything running when scale doubles overnight.
• You don't need to be world-class at all three on day one.
• But you need to be exceptional at one, curious about the others, and someone who gets uncomfortable when they don't understand how something works end-to-end.

Requirements:

• You have strong CS fundamentals and can reason from first principles, algorithms, memory, concurrency, and distributed systems, not just from framework documentation.
• You've built something non-trivial in production and cared deeply about what happened after it shipped.
• You've debugged something deeply broken and found the real root cause, not just the proximate one.
• You're genuinely curious about AI and voice systems, even if you haven't worked on them directly yet. What matters is that you learn fast and go deep.
• Experience with model training or fine-tuning is a strong plus, but not a requirement. What we care about far more is how you think, how you debug, and whether you take ownership.
• Strong backgrounds we see here: IIT/NIT/BITS graduates, engineers from Google, Nvidia, DeepMind, or fast-scaling Indian AI startups.