Manthan Gupta

Manthan Gupta works on the systems powering modern AI. His interests span agent architectures, memory systems, evals, inference engineering, and large-scale distributed infrastructure, with a particular focus on how autonomous AI systems evolve beyond simple chat interfaces. As a member of technical staff at Zyphra, he combines a researcher’s curiosity with a builder’s mindset to explore the next layer of AI infrastructure.

Gupta has built a range of projects spanning AI infrastructure, agentic systems, developer tools, and experimental programming languages. One of his notable projects is Auto-Inference-Optimiser, an autonomous agent that optimises LLM inference speed on Apple Silicon.

He also created CricLang, a dynamically typed programming language written in Go for cricket enthusiasts. The project blends sports and software engineering through cricket-themed syntax and language constructs, showcasing Gupta’s experimental approach to developer tooling.

Another major project is Water, a multi-agent orchestration framework for simplifying the creation and management of AI agent workflows. Gupta describes it as an agent harness framework built to reduce boilerplate and improve the developer experience around agentic AI systems.