Amisha Gupta

I build the machinery around AI models that keeps them reliable in production, from schema to ship.

Open to founding & product-engineering roles

Building retrieval-grounded AI systems.

Work WritingEmail

San Francisco Bay AreaUC Berkeley, Computer Science & Data Science

Product surface, backend, and the agents in between.

Amisha Gupta

Experience

A mix of YC startups, big tech, fintech, and healthcare.

Real products with real users, and the production systems underneath them.

Until they're live, recent work is on GitHub.

About

I'm Amisha. I study Computer Science and Data Science at UC Berkeley. Most of my work sits on the production side of AI systems: the parts that have to keep working after they ship.

Some of that is research. A Mixed-Integer Linear Program I wrote in Pyomo, scheduling work across federated fog systems, got published at the IEEE Cloud Summit 2025 on an NSF grant. Before that, Berkeley's Cognition and Action Lab, turning noisy human-subject data into pipelines that ran the same way twice.

Most of it is production: building and hardening backend systems and AI infrastructure across startups and larger engineering teams.

It's the same instinct in all of it. Work out the guarantee the system actually needs, then build the thing that holds it, and leave it simpler than I found it.

Education

GPA 3.875

UC Berkeley

B.A. Computer Science and B.A. Data Science, with a Computational Biology emphasis.

Abstract algebra, discrete mathematics and probability, probability theory, artificial intelligence, machine learning, deep neural networks, machine learning for bioengineering, computer security, operating systems.

Research

Cognition and Action Lab (Ivry Lab), UC Berkeley. Research assistant to Prof. Richard Ivry: human-subject motor-learning experiments and reproducible Python pipelines (NumPy, Pandas, Matplotlib) that turn noisy behavioral data into clean learning curves, reaction times, and publication-quality figures.

Skills

AI-native
MCPClaude Agent SDKOpenAI Agents SDKtool useRAGLangGraphAnthropic APIOpenRouterLlamaIndexLangChainGoogle ADKvector searchprompt engineeringworkflow orchestrationguardrailsAI observabilityfine-tuningeval pipelinesagent harnesses
Frontend
TypeScriptReactGradioStreamlitHugging Face Spaces
Backend
PythonFastAPIJavaSpring BootGoKotlinSQLPostgreSQLKafkaRedisCassandraRESTGraphQLasync pipelinesretry-aware systemscircuit breakers
Infra
AWSGCPAzureDockerKubernetesCI/CDLinuxTerraformSparkAirflowBash

Writing

All writing →

Notes on the systems I am shipping. Roughly weekly.

Let's talk.

Open to founding & product-engineering roles.

I want to work on

  • Eval pipelinesMaking non-deterministic systems measurable enough to ship on a schedule.
  • Agent infrastructureThe harness around the model: the loop, tool schemas, state, and side effects that only happen once.
  • Research-driven explorationProblems where the answer is not known yet and the work is finding the bound.
  • Backend servicesThe load-bearing parts. APIs, data pipelines, and the reliability underneath them.
GitHub LinkedIn