Startups Directory

Nscale builds and operates AI-native data centers providing vertically integrated GPU cloud infrastructure for training, fine-tuning, and running large language models. Unlike competitors reliant on expensive third-party colocation, Nscale owns and operates its own facilities, enabling cost-effective deployment of 10,000+ GPU clusters with up to 20 MW of contiguous capacity. The company offers serverless inference, managed Kubernetes orchestration, and domain-specific model fine-tuning through an API-driven platform with pay-per-use pricing.

Anysphere develops Cursor, an AI-assisted code editor built as a fork of Visual Studio Code that integrates deep codebase understanding directly into the development workflow. The platform helps software developers automate repetitive coding tasks—boilerplate, refactors, consistency maintenance—by enabling natural-language prompts instead of line-by-line typing. Unlike ChatGPT-for-coding tools, Cursor's core strength is understanding entire codebases, modules, and cross-cutting dependencies. The company's technical approach of forking VS Code rather than building a plugin enables deeper AI integration with native Git and terminal access.

Mercor operates a dual-sided marketplace connecting domain experts (engineers, lawyers, consultants, doctors) with AI labs and enterprises for specialized evaluation, testing, and training work. The platform sources, vets, and places talent on project-based roles while AI companies use contractor feedback to refine frontier models. Mercor manages over 30,000 contractors paid collectively $1.5M daily, working with top-5 AI labs including OpenAI and Anthropic, and generates revenue through a margin on successful placements.

d-Matrix develops specialized AI inference accelerator hardware for data centers, featuring the Corsair inference accelerator card, JetStream I/O networking accelerator, and SquadRack rack-scale system. The company uses digital in-memory compute architecture to deliver 3x better cost-performance and 10x faster token generation compared to traditional GPU-based inference solutions. d-Matrix targets the massive AI inference market—where most compute costs occur after model training—by optimizing for real-time latency, energy efficiency, and cost at scale.

Fireworks AI provides a unified inference and fine-tuning platform for open-source large language, vision, and multimodal models, eliminating the need for users to manage GPU infrastructure. The platform serves over 100 state-of-the-art models with 4x lower latency than competitors like vLLM and up to 12x faster inference than comparable engines. Built by former Meta and Google AI leaders, Fireworks enables developers and enterprises to deploy, customize, and scale AI models with enterprise-grade compliance and zero data retention by default.

n8n is a workflow automation platform that combines visual no-code building with code-based customization, designed for technical teams. It supports 500+ integrations, self-hosting, and native AI agent capabilities with memory, evaluations, and multi-agent orchestration. Unlike purely visual tools like Zapier, n8n includes a built-in VS Code-style code editor for JavaScript and Python, enabling complex, multi-step automations. The platform powers mission-critical workflows for 3,000+ enterprises globally.

9fin is an AI-native intelligence platform that transforms complex financial data into structured, decision-ready insights for debt market professionals. It combines proprietary data from the world's largest team of private credit journalists with AI-powered workflows and human oversight to help credit teams identify deals, analyze risk, and monitor global debt markets in minutes instead of weeks. The platform serves over 300 leading banks, asset managers, law firms, and advisory firms globally, positioning itself as a modern alternative to legacy systems like Bloomberg and FactSet with a focus on debt markets and AI automation.

Lightmatter develops photonic processors and interconnects that dramatically reduce energy consumption and latency in AI computing by replacing electrical connections with optical ones. The company's Envise chip combines traditional electronics with integrated photonic elements for AI acceleration, while its Passage interconnect technology enables data to move 100x faster between chips. They serve hyperscale cloud providers and AI data center operators, addressing the critical bottleneck of data movement energy consumption in modern AI workloads.

Sublime Security is an AI-powered email security platform that deploys autonomous agents to detect, triage, and stop targeted email attacks in real time. Built on Message Query Language (MQL), a domain-specific language for email behavior detection, it enables security teams to write and deploy custom defenses without MX record changes across Microsoft 365 or Google Workspace. The platform combines transparency and flexibility with autonomous threat investigation, reducing manual review work by 62% while catching significantly more threats than traditional blackbox solutions.

Legora is an AI-powered legal workspace that automates complex legal workflows including contract review, due diligence, document drafting, and legal research. Built with agentic AI that reasons through multi-step legal tasks, it integrates deeply into Microsoft Word and document management systems like iManage and SharePoint. The platform enables lawyers to work with AI as a co-counsel, leveraging features like Tabular Review for mass contract analysis, Research for precedent discovery, and Workflows for playbook-driven automation. Legora reduces manual document handling and accelerates legal delivery without sacrificing accuracy or oversight.

Anaconda provides a secure, governed platform for open-source Python package management and AI development across the entire lifecycle. The company offers Conda, a dependency and environment manager that ensures reproducibility and compatibility across platforms, solving critical challenges with unvetted packages, dependency conflicts, and security gaps. Trusted by 95% of Fortune 500 companies and 50 million users globally, Anaconda enables enterprises to build and deploy AI systems reliably while maintaining governance and compliance.

Decagon is an AI agent platform that enables brands to deploy personalized, omnichannel customer service agents capable of handling complex workflows like refunds, identity verification, and escalations 24/7. The platform uses a centralized intelligence layer with Agent Operating Procedures (AOPs)—plain English instructions that allow AI to learn and improve from every interaction without requiring code changes. Customers like Chime, ClassPass, and Duolingo have achieved 60-80% support cost reductions and 95% improvements in resolution efficiency. Decagon's differentiator is its ability to handle sophisticated tasks across any channel while maintaining consistency and learning continuously from customer interactions.

Deepgram is a foundational AI company that builds vertically integrated voice AI infrastructure, controlling model development, data labeling, synthetic data generation, and its own data centers. The platform offers three core capabilities: Listen (speech-to-text), Think (language understanding), and Speak (text-to-speech), with particular strength in handling real-world audio conditions like background noise, overlapping speakers, accents, and technical terminology. Serving over 1,300 organizations including NASA, Spotify, and Twilio, Deepgram differentiates through superior accuracy, lowest latency, and 2-5x more affordable pricing than competitors by owning its entire stack.

FuriosaAI designs and manufactures high-performance, energy-efficient AI accelerators (NPUs) for data center deployments of large language models, multimodal applications, and computer vision workloads. Their flagship RNGD chip delivers 2.25x better inference performance per watt compared to traditional GPUs while consuming only 180W of power. The company serves enterprise customers and hyperscalers seeking cost-effective alternatives to Nvidia GPUs for AI inference at scale.

Omni is a modern AI-native analytics platform that combines a governed semantic layer with flexible SQL and UI tools for data exploration, visualization, and reporting. It sits between raw data and users, defining metrics, access controls, and calculation rules in one place. Built by Looker and Stitch veterans, Omni resolves the traditional governance vs. agility trade-off in business intelligence, enabling self-service analytics without chaos or siloed SQL exploration.

Parloa is an AI Agent Management Platform (AMP) that enables enterprise contact centers to automate customer service at scale using generative AI agents designed through natural language briefings rather than scripts. The platform specializes in voice-first interactions across phone, chat, and messaging channels, allowing businesses to deploy teams of autonomous AI agents that engage in dynamic, natural conversations with customers. Trusted by Fortune 200 companies like Allianz, Booking.com, and SAP, Parloa has achieved $50M+ ARR with 150% net revenue retention, positioning itself as an industry-first solution for agentic AI in customer service.

Hailo develops patented silicon AI accelerators that enable efficient deep learning inference on edge devices with minimal power consumption (<3W), eliminating cloud dependency while protecting privacy. The Hailo-8 and Hailo-10H chips integrate all required memory on the processor die, delivering 26+ TOPS of performance while maintaining exceptional power efficiency. Serving automotive, security, retail, industrial, and medical sectors, Hailo has secured over 300 customers including HP, Dell, BMW, Sony, and Bosch.

Harmonic builds Aristotle, an AI-driven mathematical reasoning engine that uses formal verification via the Lean 4 proof assistant to solve complex mathematical problems with provable correctness. The platform translates natural-language math problems into formally verified proofs, eliminating the guesswork in mathematical reasoning. Harmonic serves mathematicians, quantitative researchers, software verification teams, and enterprises needing formally verified code generation. Unlike general-purpose AI models that can hallucinate intermediate steps, Aristotle guarantees each reasoning step is mathematically sound.

XBOW is an AI-powered autonomous penetration testing platform that identifies and exploits vulnerabilities in web applications without human intervention. Using swarms of AI agents, it simultaneously explores multiple attack vectors, reducing testing times from weeks to hours while providing reproducible proof of exploitation. The platform is trusted by Fortune 500 companies including UKG, Samsung SDS, and Moderna, and recently ranked #1 on the HackerOne leaderboard, outperforming thousands of human hackers.

Enfabrica develops specialized networking chips and systems that solve critical data movement bottlenecks for AI and machine learning workloads at hyperscale. Their Accelerated Compute Fabric (ACF) SuperNIC chip delivers multi-terabit Ethernet connectivity and memory disaggregation, enabling GPUs, CPUs, and memory to communicate 4x faster than competing solutions. Built by former Broadcom and Google infrastructure leaders, the company bridges PCIe/CXL memory semantics with RDMA networking to reduce AI processing costs and improve efficiency.

CAST AI is a machine learning-powered platform that automatically optimizes Kubernetes clusters and cloud infrastructure in real time, reducing cloud costs by 50% or more while improving application performance and security. Unlike competitors that merely monitor and recommend fixes, CAST AI actively implements optimizations across AWS, Azure, GCP, and on-premises environments. The platform uses advanced ML algorithms to continuously analyze workload behavior, automate resource allocation, and eliminate waste through intelligent bin packing and pod scheduling.

Recogni develops AI inference processors that deliver 10x higher compute density, 10x lower power consumption, and 13x lower cost per query compared to GPU-based solutions. The company applies proprietary logarithmic number system (LNS) technology to enable efficient inference at scale for generative AI and autonomous vehicle applications. Recogni targets cloud data centers and automotive OEMs seeking to reduce the power, cooling, and infrastructure costs of AI deployment.

FundGuard is a cloud-native, AI-powered investment accounting platform that serves as a core system of record for asset managers and fund administrators. It automates complex multi-asset accounting workflows, real-time NAV monitoring, and exception management across multiple books of record (IBOR, ABOR, X-BOR). The platform uses machine learning to automatically map financial data from any format (CSV, Excel, XML, screenshots) and transform accounting from reactive to proactive, reducing operational risk and manual effort.

You.com is an AI-powered search and agent platform that solves hallucination in large language models by grounding them with real-time, accurate web and private data. The company provides enterprise-grade search APIs and AI tools designed for AI agents and knowledge workers who need trustworthy, contextual information at scale. Founded by former Salesforce chief scientist Richard Socher, You.com has pivoted from consumer search toward B2B enterprise sales, serving finance, biotech, legal, and academic sectors with privacy-first infrastructure built for the agentic era.

Osapiens is a cloud-based, AI-powered compliance and operations automation platform that helps large enterprises manage supply chain compliance, sustainability, and regulatory requirements at scale. The platform integrates with existing ERP systems like SAP and Microsoft Dynamics, offering no-code workflow builders, real-time monitoring, and pre-built compliance modules for regulations like EUDR, CSDDD, and Germany's Supply Chain Due Diligence Act. With over 2,500 customers worldwide, Osapiens positions itself as the go-to solution for distributed organizations needing coordinated process management and audit readiness across complex operations.

Cognigy.AI is a low-code conversational AI platform that enables enterprises to automate customer and employee communications through intelligent voice- and chatbots across any channel (web, phone, SMS, mobile). The platform combines Large Language Models with structured logic workflows, allowing organizations to deploy sophisticated AI agents capable of independent decision-making while maintaining precise intent recognition. Available as SaaS or on-premise, it supports multilingual conversations and integrates seamlessly with backend systems. Now part of NICE following a $955M acquisition in September 2025.

Clay is a data enrichment and automation platform that enables sales and growth teams to find, research, and personalize outreach to leads at scale. It functions as a smart spreadsheet that orchestrates data from 150+ sources, applies AI-powered workflows for enrichment and content generation, and automates prospecting without manual labor. Clay uses signal-led prospecting to identify buyers based on real-world triggers like funding rounds or hiring, making personalized outreach possible without expensive tools or manual work.

Laurel is an AI-native time intelligence platform that passively captures digital work activity and uses firm-specific large language models to automatically generate structured, compliant time entries and billing narratives. Built specifically for enterprise professional services firms, it eliminates manual timekeeping while helping lawyers and accountants recover an average of 28 billable minutes per day—translating to 4-11% profit increases. Laurel is the first and only company applying generative AI to enterprise timekeeping, processing over $4.4 billion in professional time annually.