Founding Engineer - AI

This is a Founding Engineer position, not just a title. You'll take ownership of some of our most challenging technical problems, driving them from architecture to production with real decision-making authority from day one.

The core responsibilities for the job include the following:

LLM and ML Systems:

• Fine-tune and optimise proprietary Small Language Models (SLMs) for Indian financial conversations.
• Build training pipelines using domain-specific datasets.
• Optimise inference to achieve sub-200ms latency at scale.
• Develop intent classification and end-of-utterance detection models.

Voice and Real-Time Infrastructure:

• Push the latency limits of our real-time voice pipeline, including WebSocket audio streaming, STT, LLM, and TTS orchestration.
• Build and maintain custom telephony infrastructure capable of handling 10,000+ concurrent AI voice calls.
• Solve complex asynchronous concurrency challenges using Python.

Distributed Intelligence:

• Architect data replication pipelines, scheduling systems, and observability infrastructure.
• Build distributed systems that process millions of daily conversations.
• Work with technologies such as ClickHouse analytics, Kubernetes autoscaling, and distributed computing frameworks.

Requirements:

• Candidates from the following backgrounds are preferred: IIT, NIT, or BITS graduates.
• Engineers from Google, Nvidia, DeepMind, or fast-scaling Indian AI startups.

First-Principles Thinking:

• Strong computer science fundamentals with the ability to reason from first principles across memory management, algorithms, concurrency, and distributed systems rather than relying solely on framework documentation.

Production Ownership:

• Proven track record of building and owning non-trivial production systems.
• Strong debugging skills with the ability to identify and resolve root causes effectively.

Fast Learner:

• Passion for AI and voice technologies.
• Ability to learn quickly and take complete ownership of problems.
• Curiosity and execution are valued more than prior experience in voice systems.

Bonus Points:

• Experience with low-latency or real-time systems such as audio/video streaming, VoIP, gaming backends, and high-frequency trading (HFT).
• Hands-on experience with LLM inference optimisation, including vLLM, TGI, quantisation, and LoRA / PEFT.
• Exposure to Indic language NLP or speech systems, or experience across any stage of the LLM lifecycle, including data creation, fine-tuning, and RLHF (Reinforcement Learning from Human Feedback).
• Knowl's Tech Stack: Python, FastAPI, Open Source LLMs, PostgreSQL, Azure, AWS, ClickHouse, and Kubernetes.