evaluationcontextprotocol.io
Lightweight protocol and runtime for evaluating AI agents, with structured capture of reasoning, tool usage, and outputs for deterministic, reproducible evaluation.
Manav Anandani
Building reliable AI systems and scalable software.
I build AI systems as software systems: reliable, scalable, and designed to hold up under real use.
I study Data Science at San Jose State University, where my work centers on AI systems, infrastructure, and the design of software that remains reliable as complexity grows.
I care about the difficult layer between model capability and dependable product behavior.
Selected Work
perhapsnow.app
Behavioral timing intelligence framework that models readiness states and turns subjective decision-making into a measurable, feedback-driven system.
Human-Guided RL for CartPole
Reinforcement learning study showing how human feedback can guide policy updates, accelerate learning, and improve reward performance in CartPole.
Experience
Fydo
Built AI systems and infrastructure for Money Arena, including real-time voice interaction workflows, observability pipelines, and production reliability layers.
Lumen Technologies
Developed enterprise-scale GenAI pipelines for conversational intelligence, with a focus on summarization, subintent detection, scalable batch processing, and cloud-native reliability.
Dots & Coms
Designed an AI-powered proctoring platform using computer vision, speech analysis, and automated review workflows to improve monitoring accuracy and system scalability.
Sopra Steria
Built forecasting and analytical modeling systems that improved prediction accuracy, reduced manual effort, and translated economic data into decision-ready insight.
Notes
How I think about AI systems
Good AI systems are not just model wrappers. They need evaluation, observability, strong interfaces, and software that stays useful when the environment gets messy.
What I want to build
I am most interested in agentic software, voice interfaces, evaluation tooling, domain-specific assistants, and systems where AI changes how people reason or operate.
Current focus
Right now, I am focused on AI systems infrastructure, practical LLM products, real-time interaction, and the engineering required to make intelligent software dependable.
Elsewhere
I am also exploring startups, product ideas, and founder-minded AI applications, especially where strong engineering can create leverage far beyond the model alone.