Berkeley Xcelerator

The Launchpad for Next‑Generation AI & Agentic AI Startups

Why Berkeley Xcelerator

The Berkeley Center for Responsible, Decentralized Intelligence (RDI) is building the global hub for innovation in AI and Agentic AI. At the heart of this mission is the Berkeley Xcelerator, designed to help founders turn breakthrough ideas into scalable, venture‑backable companies that shape the future of intelligent systems.

Unlike traditional accelerators, the Berkeley Xcelerator is completely non-dilutive and free for selected startups.

110+

Global teams incubated across AI, cybersecurity, and decentralized tech

$650M+

Follow-on funding from venture and strategic investors

100+

Countries represented in founder network

2026 Spring Cohort

Narada AI

Agentic automation platform using Large Action Models to automate complex enterprise workflows end-to-end.

Hover for founders →

Narada AI

Narada is an agentic automation platform developed out of the UC Berkeley AI Research Lab that enables enterprises to automate complex, multi-step workflows across their business applications. Their AI agents execute processes end-to-end across both SaaS and legacy systems, adapt and self-heal to variable interfaces, and operate via the browser, APIs, and Citrix.

Dave Park
Dave Park — Co-Founder & CEO in

Co-founded Coverity out of Stanford ($375M acquisition by Synopsys). Serial entrepreneur.

Amir Gholami
Amir Gholami — Co-Founder & CTO in

Research Scientist at UC Berkeley BAIR and Sky lab. PhD UT Austin.

Kurt Keutzer
Kurt Keutzer — Co-Founder & CSO in

Professor, UC Berkeley EECS. Former CTO of Synopsys. Co-founded DeepScale (acq. Tesla), Nexusflow (acq. NVIDIA).

Berkeley, CA Founded 2021
www.narada.ai ↗ connect@narada.ai ✉ LinkedIn ↗

cognee

Ontology-grounded context engine for your agents, enabling them to reason and improve over time.

Hover for founders →

cognee

Open-source AI memory layer using vectors and knowledge graphs. Structures data into vectors and knowledge graphs, enabling persistent context and multi-agent coordination for AI applications.

Vasilije Markovic
Vasilije Markovic — Founder & CEO in

Prev. Zalando, Taxfix, Omio. Background in data engineering and cognitive sciences.

Berlin, Germany Founded 2024
www.cognee.ai ↗ social@cognee.ai ✉ LinkedIn ↗

Nimblemind

AI agents for healthcare data interoperability, automating feature extraction from multi-modal healthcare data.

Hover for founders →

Nimblemind

Nimblemind trains and builds agents for Healthcare Data Interoperability. Nimblemind automates feature extraction from multi-modal healthcare data.

Pi Zonooz
Pi Zonooz — Co-Founder & CEO in

Prev. AWS and Stripe. Nuclear Engineering, UIUC. MBA, Northwestern.

Navin Kumar
Navin Kumar — Co-Founder & Chief AI Officer in

PhD from Yale. Prev. Head of Data Science at NYC Health + Hospitals.

Chicago, IL Founded 2024
www.nimblemind.ai ↗ contact@nimblemind.ai ✉ LinkedIn ↗

Headroom

The context optimization layer for LLM agents, cutting tokens 40-90% losslessly across text, voice, and regulated verticals.

Hover for founders →

Headroom

Headroom is the context optimization layer for LLM agents, cutting tokens 40-90% losslessly across text, voice, and regulated verticals. The open-source proxy (1,700+ GitHub stars, 40K+ PyPI downloads, 200B+ tokens saved) integrates with LangChain, Claude Code, Codex, and MCP.

Tejas Chopra
Tejas Chopra — Founder & CEO in

Prev. Netflix (ML/AI Infrastructure), Box, Samsung, Apple. MS Carnegie Mellon.

Fremont, CA Founded 2026
headroomlabs.ai ↗ hello@headroomlabs.ai ✉

RELAI

The learning engine for self-improving AI agents.

Hover for founders →

RELAI

RELAI is a lifelong optimization engine for AI agents that turns production failures and user feedback into replayable test environments. Its continuous learning loop helps enterprise teams improve prompts, models, tools, memory and agent structure while increasing reliability and preventing regressions.

Laurent Py
Laurent Py — Co-Founder & CEO in

Multi-time founder. PhD in Computer Science. Scaled SaaS businesses up to $80M.

Soheil Feizi
Soheil Feizi — Co-Founder & CSO in

Associate Professor, University of Maryland. PhD MIT, postdoc Stanford. 2024 PECASE recipient.

Priyatham Kattakinda
Priyatham Kattakinda — Co-Founder & Head of Engineering in

PhD from University of Maryland. Prev. research in ML systems reliability.

Wenxiao Wang
Wenxiao Wang — Co-Founder & Head of AI in

PhD from University of Maryland. BS from Tsinghua University (Yao Class).

Bethesda, MD Founded 2024
relai.ai ↗ contact@relai.ai ✉ LinkedIn ↗

AgntID

Access Control Enforcement for AI Agents.

Hover for founders →

AgntID

Runtime enforcement layer for AI agent access control with scoped, ephemeral credentials per tool invocation. Intercepts agent tool calls, evaluates policy and intent, derives scoped credentials, and executes. Integrates with existing IAM and MCP ecosystems without requiring agent redesign.

Sundar Krish
Sundar Krish — Co-Founder & CEO in

Serial entrepreneur. Founded Sken.ai (acquired by Fortinet) and waygum.io (acquired).

R
Ramanuj Pandey — Co-Founder & CTO in

Prev. Fortinet, Juniper. 25+ years in security platforms.

Dublin, CA Founded 2025
agntid.ai ↗ LinkedIn ↗

Founding Dev

Replace your SaaS stack with software you own.

Hover for founders →

Founding Dev

Founding Dev helps companies replace bloated and expensive SaaS subscriptions with tools they own and control. AI agents and production-ready templates let teams deploy custom internal tools in minutes and fully own the source code, infrastructure, and data.

Talha Masood
Talha Masood — Founder & CEO in

Prev. co-founded Remotebase (backed by Tim Draper, Hustle Fund). Forbes Technology Council member.

San Francisco, CA Founded 2025
founding.dev ↗ talha@founding.dev ✉

ArmorIQ

AI agent security platform that enforces intent-based policy at the reasoning-to-execution boundary.

Hover for founders →

ArmorIQ

ArmorIQ is an AI agent security platform that acts as a control fabric between autonomous agents and enterprise governance domains, intercepting actions in real time and enforcing intent-based policy before any API call, data access, or workflow executes. Unlike traditional IAM or guardrails, ArmorIQ addresses the emerging threat of agents acting outside their delegated scope, blocking rogue behavior at the reasoning-to-execution boundary with full audit trails for SOC2, GDPR, and NIST compliance.

Ram Shanmugam
Ram Shanmugam — Co-Founder & CEO in

Three-time founder with two acquisitions (AutonomIQ by Sauce Labs, appOrbit by Cohesity).

Ketan Bhardwaj
Ketan Bhardwaj — Co-Founder & CTO in

PhD Computer Science, Georgia Tech. Senior Research Engineer at Georgia Tech.

Vaibhav Bhosale
Vaibhav Bhosale — Co-Founder & Chief Scientist in

PhD candidate, Georgia Tech. Prev. Microsoft.

San Francisco, CA Founded 2023
armoriq.ai ↗ founders@armoriq.io ✉ LinkedIn ↗

What We Offer

Unparalleled Research Access

Plug directly into Berkeley's frontier research at the intersection of Agentic AI Foundation, AI Safety & Security, and related domains, giving companies an early advantage grounded in scientific excellence.

Elite Mentorship & Advisors

Founders receive guidance from renowned faculty, serial entrepreneurs, industry leaders, and top venture investors with experience scaling successful companies.

Global Innovation Community

Startups access Berkeley RDI's network of 40,000+ builders, developers, researchers, and founders worldwide collaborating on responsible AI advancement.

Market Visibility & Validation

Companies showcase their work at summits and demo days to build credibility with investors, customers, and enterprise partners.

Cloud & GPU Resources

Participants receive cloud credits, GPU access, and computational support from our industry partners to accelerate product development.

  • Up to $10,000 in API credits from OpenAI
  • Up to $10,000 in Google Cloud and Gemini credits from Google Cloud & DeepMind
  • Up to $10,000 in compute/inference credits from Nebius
  • More cloud, GPU, and infrastructure support to be announced

What You'll Achieve

By the end of the Xcelerator cohort, teams are expected to demonstrate meaningful progress across the following areas:
  • Clearly Defined Problem & Target Customer
    A well-articulated problem statement, a clearly identified target customer profile, and a validated use case.
  • Working Product Demo
    A functional product demo (from MVP to early v1) showcasing core capabilities.
  • Early Validation & Traction
    Tangible evidence of validation, such as user engagement, pilot programs, customer feedback or other credible signals of demand.
  • Go-to-Market Strategy
    Clear plan for distribution, scaling, user/customer acquisition, and early revenue.

Our Partners

Our Partners - DeepMind, Google Cloud, NEBIUS, and more to be announced soon