Pink Hat, a supplier of open supply options, has unveiled updates to Pink Hat AI, its portfolio of services designed to assist speed up the event and deployment of AI options throughout the hybrid cloud.
Pink Hat AI offers an enterprise AI platform for mannequin coaching and inference that delivers elevated effectivity, a simplified expertise and the pliability to deploy anyplace throughout a hybrid cloud surroundings.
Whilst companies search for methods to scale back the prices of deploying giant language fashions (LLMs) at scale to deal with a rising variety of use circumstances, they’re nonetheless confronted with the problem of integrating these fashions with their proprietary information that drives these use circumstances whereas additionally with the ability to entry this information wherever it exists, whether or not in an information centre, throughout public clouds and even on the edge.
Encompassing each Pink Hat OpenShift AI and Pink Hat Enterprise Linux AI (RHEL AI), Pink Hat AI addresses these considerations by offering an enterprise AI platform that allows customers to undertake extra environment friendly and optimised fashions, tuned on business-specific information and that may then be deployed throughout the hybrid cloud for each coaching and inference on a wide-range of accelerated compute architectures.
Pink Hat OpenShift AI
Pink Hat OpenShift AI offers an entire AI platform for managing predictive and generative AI (gen AI) lifecycles throughout the hybrid cloud, together with machine studying operations (MLOps) and LLMOps capabilities. The platform offers the performance to construct predictive fashions and tune gen AI fashions, together with instruments to simplify AI mannequin administration, from information science and mannequin pipelines and mannequin monitoring to governance and extra.
Pink Hat OpenShift AI 2.18, the newest launch of the platform, provides new updates and capabilities to help Pink Hat AI’s intention of bringing higher optimised and extra environment friendly AI fashions to the hybrid cloud. Key options embody:
- Distributed serving: Delivered via the vLLM inference server, distributed serving allows IT groups to separate mannequin serving throughout a number of graphical processing models (GPUs). This helps reduce the burden on any single server, hurries up coaching and fine-tuning and makes extra environment friendly use of computing assets, all whereas serving to distribute companies throughout nodes for AI fashions.
- An end-to-end mannequin tuning expertise: Utilizing InstructLab and Pink Hat OpenShift AI information science pipelines, this new characteristic helps simplify the fine-tuning of LLMs, making them extra scalable, environment friendly and auditable in giant manufacturing environments whereas additionally delivering manageability via the Pink Hat OpenShift AI dashboard.
- AI Guardrails: Pink Hat OpenShift AI 2.18 helps enhance LLM accuracy, efficiency, latency and transparency via a know-how preview of AI Guardrails to watch and higher safeguard each person enter interactions and mannequin outputs. AI Guardrails gives further detection factors in serving to IT groups determine and mitigate probably hateful, abusive or profane speech, personally identifiable info, aggressive info or different information restricted by company insurance policies.
- Mannequin analysis: Utilizing the language mannequin analysis (lm-eval) part to supply vital info on the mannequin’s general high quality, mannequin analysis allows information scientists to benchmark the efficiency of their LLMs throughout a wide range of duties, from logical and mathematical reasoning to adversarial pure language and extra, in the end serving to to create more practical, responsive and tailor-made AI fashions.
RHEL AI
A part of the Pink Hat AI portfolio, RHEL AI is a basis mannequin platform to extra constantly develop, take a look at and run LLMs to energy enterprise functions. RHEL AI offers clients with Granite LLMs and InstructLab mannequin alignment instruments which are packaged as a bootable Pink Hat Enterprise Linux server picture and may be deployed throughout the hybrid cloud.
Launched in February 2025, RHEL 1.4 added a number of new enhancements together with:
- Granite 3.1 8B mannequin help for the newest addition to the open source-licensed Granite mannequin household. The mannequin provides multilingual help for inference and taxonomy/information customisation (developer preview) together with a 128k context window for improved summarisation outcomes and retrieval-augmented technology (RAG) duties.
- A brand new graphical person interface for abilities and information contributions, accessible as a developer preview, to simplify information ingestion and chunking in addition to how customers add their very own abilities and contributions to an AI mannequin.
- Doc Data-bench (DK-bench) for simpler comparisons of AI fashions fine-tuned on related, personal information with the efficiency of the identical un-tuned base fashions.
Pink Hat AI InstructLab on IBM Cloud
More and more, enterprises are searching for AI options that prioritise accuracy and information safety, whereas additionally protecting prices and complexity as little as potential. Pink Hat AI InstructLab deployed as a service on IBM Cloud is designed to simplify, scale and assist enhance the safety footprint for the coaching and deployment of AI fashions. By simplifying InstructLab mannequin tuning, organisations can construct extra environment friendly fashions tailor-made to the organisations’ distinctive wants whereas retaining management of their information.
No-cost AI Foundations coaching
AI is a transformative alternative that’s redefining how enterprises function and compete. To help organisations on this dynamic panorama, Pink Hat now gives AI Foundations on-line coaching programs without charge. Pink Hat is offering two AI studying certificates which are designed for knowledgeable senior leaders and AI novices alike, serving to educate customers of all ranges on how AI may also help remodel enterprise operations, streamline decision-making and drive innovation. The AI Foundations coaching guides customers on how one can apply this data when utilizing Pink Hat AI.
Pink Hat AI InstructLab on IBM Cloud will probably be accessible quickly. AI Foundations coaching from Pink Hat is on the market to clients now.
Joe Fernandes, VP and normal supervisor, AI Enterprise Unit, Pink Hat, mentioned: “Pink Hat is aware of that enterprises will want methods to handle the rising value of their generative AI deployments, as they bring about extra use circumstances to manufacturing and run at scale. In addition they want to deal with the problem of integrating AI fashions with personal enterprise information and be capable to deploy these fashions wherever their information could dwell. Pink Hat AI helps enterprises handle these challenges by enabling them to leverage extra environment friendly, purpose-built fashions, skilled on their information and allow versatile inference throughout on-premises, cloud and edge environments.”
Régis Lesbarreres, superior analytics and AI innovation supervisor, digital innovation, Airbus Helicopters, mentioned: “On the outset of our AI journey, Airbus Helicopters was seeking to combine AI into our current structure, cut back shadow IT, unite our information scientists beneath a singular AI platform and optimise prices at scale. With Pink Hat OpenShift AI, we’ve been in a position to accomplish all of those objectives, which led to our first enterprise use case for AI. Pink Hat’s imaginative and prescient for AI aligns with our enterprise goals and permits us to fulfill them whereas sustaining flexibility, accessibility and transparency.”
Javier Olaizola Casin, international managing accomplice, hybrid cloud and information, IBM Consulting, mentioned: “Companies are more and more making use of AI to rework core enterprise processes, they usually want AI options which are versatile, cost-effective and tuned with trusted enterprise information to fulfill their distinctive wants. Pink Hat AI brings the consistency, reliability and velocity that organisations want to construct and deploy AI fashions and functions throughout hybrid cloud situations. Combining IBM Consulting’s area, know-how and trade experience with Pink Hat’s AI applied sciences, we’re serving to our shoppers drive ROI from their know-how investments.”
Torsten Volks, principal analyst, utility modernisation, ESG, mentioned: “Main organisations harness AI-driven, data-centric resolution making throughout groups and enterprise models. Subsequently, the power to quickly and affordably develop, deploy, combine, scale and govern AI-based capabilities throughout the enterprise turns into a essential success issue. Establishing this functionality requires an open and extensible AI basis that ensures seamless integration with current methods and processes, operational agility and steady governance. Enabling workers and clients to profit from AI capabilities sooner and in a extra complete method is essential for continued enterprise success.”
Anand Swamy, govt VP and international head of ecosystems, HCLTech, mentioned: “With a purpose to realise the total potential of generative AI, organisations have to prioritise agile and versatile infrastructure. By combining the capabilities of Pink Hat AI, encompassing RHEL AI and Pink Hat OpenShift AI to ship an end-to-end AI utility platform, with HCLTech’s main AI experience and cognitive infrastructure companies, part of HCLTech AI Foundry answer, clients get a streamlined path to unlocking AI innovation, serving to them overcome widespread challenges corresponding to information safety, scaling AI workloads and minimising infrastructure prices.”
Wish to study extra about cybersecurity and the cloud from trade leaders? Take a look at Cyber Safety & Cloud Expo going down in Amsterdam, California, and London.
Discover different upcoming enterprise know-how occasions and webinars powered by TechForge right here.
