Labelbox

Tabnine, founded in 2018, develops an AI code assistant that provides intelligent code completion for software developers. The product adapts to individual coding styles and team standards, and integrates directly into developers' existing IDEs. It supports multiple programming languages and can be deployed in the cloud, on-premises, or in air-gapped environments - a flexibility that addresses enterprise requirements around data control and regulatory compliance. The company's technical work spans code-native large language models (LLMs), multi-agent systems, and IDE integrations. Privacy, security, and compliance are treated as core product attributes rather than add-ons, reflecting a design philosophy that keeps code within the developer's chosen environment. More than one million developers use Tabnine daily. Tabnine operates as a remote-first, distributed organisation. Its teams include engineers and researchers focused on applied AI and developer productivity tooling. The company frames its mission around augmenting developer creativity and autonomy, with an emphasis on building tools that respect code ownership and work within established team workflows.

Encord builds data infrastructure and tooling for AI and machine learning teams. Its platform covers the full data development lifecycle: managing and curating datasets, annotating data at scale with workforce management features, and evaluating models in production for reliability and correct behaviour. The company positions data quality as the central variable in whether AI systems perform as intended. The platform is designed for multimodal AI workflows, supporting teams that need to move from raw data to production-ready models. Core capabilities include: Data management and curation – tools for organising, filtering, and maintaining the datasets used to train and validate models Annotation and workforce management – labeling infrastructure and tooling to coordinate annotation work at scale Model evaluation and observability – capabilities for assessing model behaviour both during development and in production Encord was founded by former quantitative analysts, physicists, and computer scientists, and its team includes engineers and operators who have worked at Meta, Microsoft, Apple, Intel, Goldman Sachs, McKinsey, and J.P. Morgan. The company is backed by Y Combinator, Next47, and CRV, among others.

Explorium operates a B2B data layer designed to supply the external data and infrastructure that AI-powered products and go-to-market agents need to function effectively. Its platform aggregates and harmonizes data from more than 50 trusted sources, providing coverage of 146 million companies and 767 million professional profiles. Delivery options include API and MCP integrations, making it straightforward to embed Explorium's data into agent workflows and AI pipelines. The platform is built around several capabilities relevant to GTM and AI use cases: structured and harmonized data outputs, natural language interfaces that simplify querying, automated target discovery, and personalized signal generation. The underlying data infrastructure is designed to handle the reliability and consistency requirements of enterprise-grade deployments. Explorium's external data platform is used to accelerate agent development, enhance contextual intelligence in AI products, and support B2B data enrichment initiatives. Customers include Taboola, PepsiCo, Outreach, Bombora, and Deloitte - spanning media, consumer goods, sales technology, intent data, and professional services.

Unqork is an AI-first, no-code enterprise application platform designed to replace traditional coding with visual, component-based development. The platform targets large organisations seeking to build new applications or modernise existing systems without accumulating technical debt. Its primary industry focus spans insurance and underwriting, government and public sector, and broader enterprise IT modernisation. The platform's core offering centres on a visual, component-based environment where applications are assembled rather than written. Unqork claims this approach eliminates technical debt at the architectural level, producing systems that improve over time rather than degrade. Among its more specialised products is an AI-powered underwriting workbench, aimed at streamlining insurance workflows. The company also works on government IT transformation projects, with a stated focus on modernising critical public-sector systems. Unqork describes its engineering culture as collaborative and non-hierarchical. The company explicitly rejects what it calls a "superhero culture," instead emphasising cross-functional teamwork, shared ownership of outcomes, and open communication. Innovation is said to come from diverse perspectives and employees working outside their strictly defined roles. The team characterises its mission as delivering on commitments to customers rather than simply shipping software.

Duku AI builds autonomous testing infrastructure for software teams. Its platform runs every critical user journey after each build, simulating real user behaviour to catch failures before they reach production. Tests self-heal as the codebase evolves, with the platform designed to eliminate both flaky tests and the manual maintenance burden that typically accompanies them. The company is venture-backed and reported $330,000 in revenue in 2025, generated by a three-person team. Its founders claim operational backgrounds that include scaling Meta's testing infrastructure, launching Uber's global playbooks, and growing Deliveroo from its early stages through hypergrowth. Duku AI operates in the software quality and QA space, with its platform aimed at engineering teams, product teams, and QA functions that rely on continuous testing as part of their build and release cycles. The platform's autonomous agents are designed to reason about test failures and apply fixes without manual intervention.

In 2019, Edo Liberty was working as a research director at AWS and at Yahoo before that, where he witnessed the tremendous power of combining AI models and vector search to dramatically improve applications such as spam detectors and recommendation systems. While building custom vector search systems at enormous scales, he assumed there was already a packaged solution available for everyone else who didn't have access to the same engineering and data-science resources. To his surprise, there wasn't. This gap in the market led Edo to found Pinecone, creating the vector database category of solutions from scratch. The company was born from the recognition that developers and engineering teams of all sizes needed accessible storage and retrieval infrastructure for building and running state-of-the-art AI applications. This founding principle of accessibility drove Pinecone's evolution into a fully managed service known for its ease of use, enabling thousands of companies to ship AI applications faster and more confidently.