Open to opportunities

Kartikey Sondhi

@kartikeysondhi

Message

I build production AI systems, retrieval pipelines, and low-latency backend infrastructure.

Zimbabwe

Message

What I'm looking for

I'm looking to own ambiguous AI and backend problems end-to-end, building reliable low-latency production systems with measurable impact and simple architectures.

At Deloitte, I own production AI applications from ambiguous business requirements through architecture, APIs, retrieval, evaluation, deployment, and production support.

I've redesigned asynchronous and parallel workflows with Celery and Redis, reducing end-to-end latency by approximately 90% while improving throughput. I also simplified architecture and operations enough to reduce system ownership from four engineers to one.

I'm building Interro, a fully local real-time voice AI interrogation game that combines speech processing, language models, agent logic, and local inference. My work centers on measurable quality, including retrieval systems with 96% precision, 18% higher answer accuracy, and 94% semantic query re-framing accuracy.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist 2

Current

Deloitte

Jul 2025 - Present (1 year 1 month)

Owns the full lifecycle of production AI applications, translating ambiguous business requirements into technical architectures, APIs, retrieval systems, evaluation pipelines, and deployed solutions. Reduced end-to-end latency by ~90% by redesigning execution around asynchronous processing, caching, batching, and parallel workflows.