Startups Directory

Reflection AI is building frontier AI models and reasoning systems designed to act as an intelligent middle layer between developers and fully autonomous systems. The company positions itself as both an open-source alternative to closed frontier labs like OpenAI and Anthropic, and a Western equivalent to Chinese AI firms like DeepSeek. Their self-hosted AI agents can reason about code, synthesize features, refactor, debug, and maintain systems while dynamically referencing files and discussions with line-level citations. Enterprise customers deploy Reflection's systems within their own cloud environments (AWS, Azure, GCP) with full privacy and IAM compliance.

Saviynt's Identity Cloud is an AI-powered platform that maps, governs, and secures all types of identities—human, machine, and AI agents—across on-premises, cloud, and hybrid environments. The platform unifies identity governance and administration (IGA), privileged access management (PAM), application access governance, and identity security posture management into a single SaaS solution. It helps enterprises automate lifecycle tasks like onboarding, offboarding, access reviews, and compliance reporting while enforcing just-in-time security principles. Saviynt serves 600+ enterprises including more than one-fifth of the Fortune 100, managing over 100 million identities at half the cost of competitors like SailPoint.

Aleph Alpha develops Luminous, a multimodal and multilingual large language model with industry-first explainability features that make AI decision-making transparent and controllable. The company also offers Pharia AI, an end-to-end generative AI stack for enterprise applications. Built on a tokenizer-free architecture, their models address the 'black box' problem of generative AI while enabling superior fine-tuning across languages and specialized industries. They serve only enterprises and governmental agencies, positioning themselves around compliance, explainability, and technological independence from U.S. vendors.

Nexthop AI builds custom networking infrastructure purpose-built for AI workloads in hyperscale cloud data centers. The company combines custom hardware, software, and optical integration using a JDM (Joint Development Manufacturer) model to deliver tightly integrated solutions for GPU clusters. Nexthop's Disaggregated Spine architecture achieves up to 30% reductions in energy consumption and infrastructure costs while supporting extreme throughput (50-100+ terabits) compared to traditional networking vendors like Cisco and Juniper.

Poolside is an enterprise-grade AI coding assistant that generates, tests, and refactors code using natural language prompts. The platform includes Malibu (a high-capacity model for complex tasks) and Point (a lightweight model for sub-200ms code completion), accessible via IDE plugins and CLI. Built by former GitHub leadership, Poolside differentiates through proprietary RLCEF technology that generates synthetic training data, enabling continuous model improvement without relying on customer data, which remains on-premise.

Adept is an enterprise AI platform that automates complex, multi-step workflows across software applications by directly perceiving and acting on user interfaces like a human would. It uses proprietary models to locate elements on webpages, reason about documents, and execute end-to-end business processes—from data extraction and form population to license application processing. The platform is designed for enterprises seeking to offload repetitive tasks while keeping humans in the loop for critical decisions.

Together AI is an AI Acceleration Cloud platform enabling developers and researchers to train, fine-tune, and deploy generative AI models at scale. The platform supports over 200 open-source models across all modalities (chat, image, audio, vision, code, embeddings) with 2-3x faster inference than hyperscalers, OpenAI-compatible endpoints, and infrastructure scaling from 16 to 100,000+ GPUs. Founded by Stanford professors and serial entrepreneurs, Together AI is positioned as the cloud infrastructure layer for the open-source AI ecosystem, with recent expansion into data transformation through the Refuel.ai acquisition.

Black Forest Labs builds FLUX, a suite of state-of-the-art generative AI models for hyperrealistic image generation and editing from text prompts. The company serves enterprise creative platforms, content creators, and developers through both a managed API and open-source model releases. FLUX models deliver exceptional prompt adherence, character consistency, and sub-second latency at production scale, with editing capabilities (Fill, Depth, Canny, Redux) enabling precise, non-destructive visual control. Notable integrations include Adobe Creative Cloud, Meta, Microsoft, Canva, and xAI's Grok.

Genspark is an AI super agent platform that automates complex, multi-step tasks by orchestrating 30+ foundation models, 150+ in-house tools, and 20 premium datasets to deliver finished outputs like presentations, videos, code, and real-world actions. It serves knowledge workers and enterprises seeking end-to-end task autonomy rather than step-by-step guidance. The platform bundles research, content creation, data analysis, and task automation into a unified workspace, differentiating itself from point-solution AI tools like ChatGPT and Perplexity through its integrated, all-in-one approach.

Positron AI designs purpose-built FPGA-based hardware (Atlas) optimized for generative AI inference workloads, achieving 93% memory bandwidth utilization versus 10-30% in GPU systems while consuming less than a third of the power of Nvidia H100s. The company serves enterprises and cloud providers running transformer-based models at scale, from content moderation to token-as-a-service platforms. Their next-generation chip (Asimov/Titan) targets memory-intensive applications like video, trading, and multi-trillion parameter models with 6x more RAM than competing solutions.

Multiverse Computing develops compressed AI models and quantum software that enables enterprises to run large language models 4x-12x faster with 50-80% lower inference costs. Using tensor network mathematics from quantum mechanics, their CompactifAI platform compresses LLMs by up to 95% with minimal precision loss, while their Singularity platform allows organizations to leverage quantum computing for optimization without specialized hardware. Serving over 100 global customers including Iberdrola, Bosch, and the Bank of Canada, they're positioned as Europe's largest quantum software company focused on practical, cost-effective AI and quantum solutions for manufacturing, finance, energy, and defense sectors.

Imbue builds foundation models optimized for reasoning to power robust AI agents that can code, understand goals, and work autonomously. The company takes a full-stack approach—training custom large language models (>100B parameters), developing agentic systems, and building infrastructure—rather than relying on third-party models. With $232M in funding and a $150M Dell partnership for a dedicated 10,000 H100 GPU cluster, Imbue is positioning itself as an independent player in AI agent development, not competing directly with foundation model companies but complementing the broader AI ecosystem.

Distyl AI is an AI-native transformation consultant that converts business processes and SOPs into auditable AI workflows using its proprietary Distillery platform. It combines software subscriptions with high-touch professional services, deploying embedded engineering teams directly with Fortune 500 clients to implement enterprise-scale AI systems. The company operates as a premium consulting partner rather than a standalone SaaS vendor, bundling platform access with multi-year, outcome-based contracts worth tens of millions of dollars.

Xelix is an AI-powered accounts payable intelligence platform that bolts onto existing ERP and AP systems to enhance financial controls and prevent fraud. Unlike traditional AP automation vendors, Xelix focuses on assurance, overpayment detection, and vendor statement reconciliation using proprietary AI models trained on seven years of data. The platform processes $200B+ in annual invoice flow, reviewing 500+ data points per invoice to identify threats and errors in real-time, and has achieved a 98% customer retention rate.

Goodfire is an AI interpretability platform that enables researchers and engineers to understand and control the internal mechanisms of large language models like Llama 3.3 70B. The company offers Ember, a hosted API for mechanistic interpretability research that lets users shape model behavior by directly controlling internal features, and Silico, a tool for adjusting model parameters during training. By reverse-engineering foundation models through Sparse Auto Encoders, Goodfire has demonstrated real-world impact—reducing hallucinations by 58% at ~90x lower cost than traditional methods—making it the first commercial product in the mechanistic interpretability space.

PaleBlueDot AI operates a dual-model platform connecting AI startups and enterprises with GPU compute capacity through a real-time marketplace and custom cluster design services. The company aggregates spare GPU capacity from third-party providers and designs large-scale clusters deployed in data centers like Digital Reality Trust and Equinix. With over 80 global clusters, it serves customers across Japan, South Korea, Singapore, and Southeast Asia—including RedNote (Xiaohongshu)—positioning itself as a cost-efficient alternative to hyperscalers for AI workloads.

PhysicsX is a physics AI platform that accelerates engineering and manufacturing by combining AI-driven multiphysics inference with numerical simulation. The platform reduces simulation run times from hours to seconds and enables engineers to run 100,000 simulations in a day versus weeks with traditional methods. Built by former Formula 1 engineers, it integrates with industry-standard CAE tools like ANSYS, CATIA, and Siemens NX, spanning design, manufacturing, and operations across multi-physics domains.

Angle Health is an AI-enabled health insurance platform that combines carrier, third-party administrator, and managing general underwriter capabilities into a single integrated system. It enables brokers to generate firm underwritten quotes on custom group health plans in minutes instead of weeks, while using proprietary AI models trained on millions of patient records to predict healthcare risks and personalize interventions. The platform integrates with 100+ payroll and HRIS systems, serving 2,600+ small business employers across 44 states with plan designs ranging from HDHPs to comprehensive PPO-style options.

Enfabrica is a fabless networking semiconductor company that designs ultra-high-performance chips for AI infrastructure. Their flagship Accelerated Compute Fabric SuperNIC (ACF-S) delivers 3.2 terabits/second of connectivity between GPUs, CPUs, and accelerators in data centers, reducing GPU compute requirements by ~50% at equivalent performance. The company also offers EMFASYS, a system combining RDMA Ethernet networking with CXL-based memory to provide elastic memory capacity for AI inference workloads. Built by veterans from Broadcom and Google, Enfabrica differentiates through vendor-agnostic architecture that avoids proprietary lock-in.

Code Metal is an AI-powered development platform that automatically translates high-level research code (Python, Julia, MATLAB) into optimized, production-ready code for any hardware target (CPUs, GPUs, FPGAs). It uses formal methods to mathematically verify translation accuracy, ensuring compliance and correctness for mission-critical systems. The platform enables engineers to write once and deploy across multiple languages and hardware platforms while maintaining code safety guarantees that LLMs alone cannot provide.

Reka is an AI research and product company that develops multimodal AI models capable of understanding text, images, video, and audio natively. The company builds frontier AI models that match or exceed competitors like GPT-4 and Gemini Ultra while requiring significantly less computational power. Reka serves enterprises and government organizations through products like Reka Core (multimodal model), Reka Vision (video/image analysis platform), and Reka Research (agentic AI for information synthesis). What differentiates them is their focus on efficiency—achieving state-of-the-art performance with orders of magnitude lower compute demands than competitors.

InstaDeep builds AI-powered decision-making systems that solve complex industrial problems across logistics, pharmaceuticals, and infrastructure. The company combines deep reinforcement learning and GPU-accelerated computing with productized innovation—translating research into enterprise-grade platforms like DeepPack (cargo planning), DeepChain (protein design), and chip design optimization. They serve Fortune 500 enterprises and global organizations including BioNTech, Deutsche Bahn, and the UN's FAO, differentiating through proven commercial deployments and strategic partnerships that turn customers into investors.

Nfinite is an AI-powered SaaS e-merchandising platform that enables retailers to create, manage, and display unlimited photorealistic product visuals through 3D CGI technology. The platform allows brands to generate product imagery from any angle, distance, or lighting configuration without expensive traditional photoshoots—reducing costs by up to 7x while accelerating time to market. Nfinite serves major retailers like Lowe's, Staples, and Wayfair, driving significant improvements in conversion rates (up to 53% lift) and customer engagement.

Typeface is an enterprise AI platform that generates on-brand, personalized content at scale across email, social, and web channels. It solves the content paradox by combining brand governance with high-velocity content creation, using dedicated AI models trained on each customer's brand guidelines, assets, and voice. The platform recently launched its Marketing Orchestration Engine, enabling governed, repeatable cross-channel campaign execution. It's backed by $206M in funding and counts Fortune 500 companies and digital-native brands among its customers.

Quincus is an enterprise SaaS platform that optimizes supply chain operations through AI-driven multi-modal logistics orchestration and real-time visibility. The platform serves e-commerce, logistics, and freight companies across Asia-Pacific, Americas, and Europe with machine learning-powered shipment allocation and a centralized control tower for end-to-end tracking. The company differentiates through advanced AI optimization that reduces fleet requirements by 33% while increasing stops per vehicle by 50%, and currently processes over 70 million shipments monthly.

Pryon is an AI Memory Layer that transforms unstructured enterprise data into searchable, conversational interfaces. The platform ingests multimodal content (text, audio, images, video) from dozens of sources like SharePoint, Salesforce, and Google Drive, then deploys proprietary AI models to provide accurate, grounded answers without hallucinations. It enables employees to instantly find and leverage knowledge scattered across the organization, reducing search time and unleashing unused enterprise content.

Inspiren is an AI-powered platform that unifies resident safety, care planning, staffing, and emergency response for senior living communities. The system combines proprietary hardware (the AUGi care companion device using geometric exoskeletal monitoring) with integrated software to detect falls, behavioral risks, and care utilization patterns in real-time while preserving resident privacy. Deployed across 230+ senior living communities, Inspiren has demonstrated measurable clinical outcomes including fall reduction and timely emergency response, with customers like Arrow Senior Living and Heritage Communities achieving 86% reductions in bedroom-related falls with injury.

Decart is an AI research lab building the fastest and most efficient generative AI models for real-time video generation and manipulation. Their flagship products—Oasis (real-time AI game world) and Mirage (live video editing via text prompts)—achieve 10x better efficiency than competitors like OpenAI's Sora, reducing inference costs from $100/hour to $0.25/hour. The company combines a proprietary LLM inference engine (built from scratch in C++ and CUDA) with advanced video generation capabilities, serving developers, enterprises, and streaming/gaming platforms.

DataSnipper is an intelligent automation platform that embeds directly into Excel, allowing auditors to extract data from unstructured sources (invoices, bank statements, documents) with automatic reconciliation and complete audit trails. Used by 600K+ professionals across all Big Four accounting firms and 175+ countries, it solves the manual, error-prone work of matching and verifying financial data. The platform's key differentiator is working within Excel workflows rather than forcing migration to new systems, enabling rapid adoption and high user loyalty.

EnCharge AI develops analog in-memory computing hardware that performs AI computations directly within memory cells, eliminating energy-intensive data movement between processing and storage. The company's EN100 accelerator delivers 200+ TOPS of AI compute in an 8.25W power envelope for laptops and workstations, achieving up to 20x better performance per watt than competing solutions. Available in M.2 and PCIe form factors, EN100 brings advanced AI capabilities to edge devices and client platforms with unprecedented efficiency.

Basis builds AI agents that autonomously execute complex accounting workflows end-to-end, including tax return preparation, audit testing, and reconciliation tasks. The platform ingests data from accounting systems and ERPs, performs analysis and calculations, and delivers finished work for human review. Top accounting firms use Basis to shift their teams from execution to review, achieving 20-50% efficiency gains across tax, audit, and advisory practices.

Hammerspace is a metadata-driven data orchestration platform that sits above file and object storage systems, providing a unified global namespace for accessing unstructured data across on-premises and public cloud infrastructure. Rather than storing data itself, it acts as a software layer that enables enterprises to access files regardless of storage type, vendor, or location while optimizing performance for AI and HPC workloads. The platform automatically transitions data between high-performance Tier 0 storage during compute-intensive operations and cost-efficient object storage during idle periods, eliminating GPU idle time caused by data bottlenecks.

Dream Security is an AI-native cybersecurity platform purpose-built for governments and critical infrastructure operators requiring sovereign deployment capabilities. The company's proprietary cyber language model—trained on threat intelligence datasets rather than general-purpose LLMs—automates threat detection, prediction, and response while operating entirely within national infrastructure, including classified and air-gapped systems. Dream transforms raw data into actionable threat narratives and defense strategies, with a founding team combining Israeli cyber expertise (NSO Group co-founder), Austrian political leadership, and IoT intelligence experience.

ZaiNar provides real-time sub-meter positioning and location intelligence using patented phase-based signal processing and sub-nanosecond time synchronization on existing radio infrastructure. The system achieves sub-10cm accuracy indoors, outdoors, and through walls without GPS, operating system permissions, or device manufacturer cooperation. It serves enterprises and carriers across healthcare, construction, smart cities, and industrial sectors, with over $450M in contracts and deployments on multiple continents.

PhotoRoom is an AI-powered product photography platform that automates image editing for e-commerce sellers. It uses deep learning to remove backgrounds, generate AI backgrounds and shadows, expand scenes, upscale images, and create product photos from scratch—all without requiring Photoshop expertise. The platform serves e-commerce businesses, marketplaces, and creative teams at scale, with APIs integrated by Shopify, Netflix, and Warner Bros. PhotoRoom's focus on simplicity and speed distinguishes it from broader photo editing tools like Photoshop.