AI startup Papaya joins Y Combinator Fall 2026 batch

Papaya, a San Francisco-based startup focused on optimizing artificial intelligence agents, has joined Y Combinator’s Fall 2026 batch as it develops technology aimed at helping companies make AI-powered workflows more efficient, less expensive and faster.

The startup was founded by Macy Mody, who serves as chief executive officer, Niyaz Puzhikkunnath, chief product officer, and Faiz Vadakkumpadath. Y Combinator lists Papaya as an active company in its Fall 2026 cohort.

Papaya’s platform is designed to examine the data produced when AI agents operate in real-world production environments. The system analyzes production traces to identify issues that may affect an agent’s accuracy, speed or operating costs.

Instead of requiring engineers to manually inspect large volumes of trace data, Papaya aims to automate the process and identify areas where an AI workflow could be improved. The platform can generate recommendations and, after engineers approve them, convert those changes into pull requests.

According to Y Combinator, Papaya connects to AI agents through a software development kit and evaluates several components of an agent’s workflow. These include the context provided to models, prompts, prompt caching, subagents and tool calls.

The company says its system conducts more than 200 research-backed analyses and ranks potential improvements according to their expected impact. Recommendations are also linked to the production runs that generated them, giving engineering teams information they can use to evaluate the suggested changes.

Papaya's approach focuses on optimizing AI agents after they have been deployed rather than relying only on testing before launch. This reflects a growing challenge for businesses as AI agents move from experimental projects into production environments and begin handling increasingly complex tasks.

Mody said the startup is focused on helping engineering teams understand why an AI agent may become slower, more expensive or less accurate once it is operating in production. Papaya is designed to identify these problems automatically and provide engineers with potential solutions.

The founders recently announced the company's acceptance into Y Combinator through a LinkedIn post, describing their recent work building the startup together before making the news public.

Y Combinator categorizes Papaya across AIOps, infrastructure and enterprise software, reflecting its focus on continuous monitoring and optimization of AI-powered systems.

Among the problems the platform can examine are oversized or inefficient context, excessive use of subagents and failed or ineffective tool calls. These issues can affect both the quality of an agent’s output and the resources required to complete a task.

Papaya says its initial workflow analyses have typically produced quality improvements of 10% or more, according to the company's Y Combinator profile. The figure represents the startup's reported results and may vary depending on the workflow and production environment being analyzed.

The founders bring experience from technology and machine learning companies. Mody previously worked in go-to-market and operations at SafeBase, where she helped the company from its early stage through its acquisition, according to Y Combinator.

Puzhikkunnath previously worked at Amazon as a senior software development engineer focused on machine learning systems. Vadakkumpadath also spent roughly a decade at Amazon, working as a senior data engineer, according to the YC profile.

Papaya is entering the market as companies increasingly experiment with AI agents that can interact with software applications, internal databases and external tools. Unlike traditional AI applications that may perform a single task, agents can carry out multiple steps and make repeated model and tool calls.

That flexibility can create new operational challenges. An agent may consume more computing resources than expected, make unnecessary tool calls, generate excessive context or take longer to complete a workflow.

For businesses deploying agents at scale, those inefficiencies can translate into higher infrastructure and model costs. They can also affect response times and the reliability of applications that depend on AI-powered workflows.

Papaya's production-focused approach is intended to give engineering teams a continuous view of how their agents perform after deployment. By analyzing actual production traces, the startup aims to identify optimization opportunities based on real-world usage rather than assumptions made during development.

The company is now developing its platform as part of Y Combinator's Fall 2026 program. Its participation gives Papaya access to the accelerator's network and resources as it works toward expanding its AI agent optimization technology for enterprise customers.

As more businesses deploy autonomous and semi-autonomous AI systems, tools designed to monitor, analyze and improve those systems could become an increasingly important part of enterprise AI infrastructure.

Papaya's entry into Y Combinator's latest cohort highlights the growing startup activity around the operational challenges created by AI agents, particularly the need to balance performance, reliability, speed and cost once these systems are running in production.