Rob Clark, President and CTO of Seekr – Interview Collection

Rob Clark is the President and Chief Know-how Officer (CTO) of Seekr. Rob has over 20 years of expertise in software program engineering, product administration, operations, and the event of modern synthetic intelligence and web-scale applied sciences. Earlier than becoming a member of Seekr, he led a number of synthetic intelligence and search options for among the world’s largest telecommunications, publishing, and e-commerce firms.

Seekr a synthetic intelligence firm, creates reliable giant language fashions (LLMs) that establish, rating, and generate dependable content material at scale.

Starting with internet search, Seekr’s ongoing innovation has produced patented applied sciences enhancing internet security and worth. Their fashions, developed with professional human enter and explainability, handle buyer wants in content material analysis, generative AI, and reliable LLM coaching and validation.

Are you able to describe the core mission of Seekr and the way it goals to distinguish itself within the aggressive AI panorama?

We based Seekr on the easy but vital precept that everybody ought to have entry to dependable, credible and reliable info – regardless of its kind or the place it exists. It doesn’t matter if you’re a web based client or a enterprise utilizing that info to make key choices – accountable AI programs permit all of us to totally and higher perceive info, as you could guarantee what’s popping out of Generative AI is correct and dependable.

When info is unreliable it runs the entire spectrum, for instance from extra benign sensationalism to the more severe case of coordinated inauthentic behaviors meant to mislead or affect as an alternative of inform. Seekr’s method to AI is to make sure the consumer has full transparency into the content material together with provenance, lineage and objectivity, and the flexibility to construct and leverage AI that’s clear, reliable, options explainability and has all of the guardrails so shoppers and companies alike can belief it.

Along with offering trade optimized Massive Language Fashions (LLMs), Seekr has began constructing Basis Fashions differentiated by higher transparency and accuracy with decreased error and bias, together with all of the validation instruments. That is made attainable by Seekr’s collaboration with Intel to make use of its newest technology Gaudi AI accelerators at the very best price-performance. We selected to not depend on outdoors basis fashions the place the coaching information was unknown and confirmed errors and inherent bias, particularly as they’re usually educated on well-liked information relatively than essentially the most credible and dependable information. We count on to launch these in the direction of the tip of the yr.

Our core product is named SeekrFlow, a whole end-to-end platform that trains, validates, deploys and scales reliable AI. It permits enterprises to leverage their information securely, to quickly develop AI they will depend on optimized for his or her trade.

What are the crucial options of SeekrFlow, and the way does it simplify the AI growth and deployment course of for enterprises?

SeekrFlow takes a top-down, consequence first method, permitting enterprises to resolve issues utilizing AI with each operational efficiencies and new income alternatives by one cohesive platform. This integration contains safe information entry, automated information preparation, mannequin fine-tuning inference, guardrails, validation instruments, and scaling, eliminating the necessity for a number of disparate instruments and decreasing the complexity of in-house technical expertise managing numerous features of AI growth individually.

For enterprises, customization is essential, as a one mannequin suits all method doesn’t resolve distinctive enterprise issues. SeekrFlow permits prospects to each cost-effectively leverage their enterprise information safely and align to their trade’s particular wants. That is particularly vital in regulated industries like finance, healthcare and authorities.

Seekr’s AI assisted coaching method enormously reduces the fee, time, and want for human supervision related to information labeling and acquisition, by synthesizing high-quality and domain-specific information utilizing domain-specific ideas equivalent to coverage paperwork, tips, or user-provided enterprise information.

Seekr’s dedication to reliability and explainability is engrained all through SeekrFlow. No enterprise desires to deploy a mannequin to manufacturing and discover out its hallucinating, giving out incorrect info or a worst-case situation equivalent to giving freely its services and products without cost! SeekrFlow contains the instruments wanted to validate fashions for reliability, to scale back errors and to transparently permit the consumer to see what’s impacting a mannequin’s output proper again to the unique coaching information. In the identical method software program engineers and QA can scan, check and validate their code, we offer the identical capabilities for AI fashions.

That is all offered at optimum value to enterprises. Our Intel collaboration operating in Intel’s AI cloud permits Seekr the perfect price-performance that we move on to our prospects.

How does SeekrFlow handle widespread points equivalent to value, complexity, accuracy, and explainability in AI adoption?

Excessive worth factors and shortage of AI {hardware} are two of the biggest limitations to entry dealing with enterprises. Due to the aforementioned collaboration with Intel, Seekr circulation has entry to huge quantities of subsequent technology AI {hardware} leveraging Intel’s AI Cloud. This supplies prospects with scalable and cost-effective compute assets that may deal with large-scale AI workloads leveraging each Intel Gaudi AI accelerators and AI optimized Xeon CPUs.

It’s vital to notice that SeekrFlow is cloud supplier and platform agnostic and runs on newest technology AI chips from Intel, Nvidia and past. Our purpose is to summary the complexity of the AI {hardware} and keep away from vendor lock-in, whereas nonetheless unlocking all of the distinctive worth of every of the chip’s software program, instruments and ecosystem. This contains each operating within the cloud or on-premise and operated datacenters.

When constructing SeekrFlow we clearly noticed the dearth of contestability that existed in different instruments. Contestability is paramount at Seekr, as we need to make sure that the consumer has the suitable to say one thing just isn’t correct and have a simple solution to resolve it. With different fashions and platforms, it’s usually troublesome or unknown easy methods to even right errors. Level fixes after the very fact are sometimes ineffective, for instance manipulating the enter immediate doesn’t assure the reply might be corrected each time or in each situation. We give the consumer all of the instruments for transparency, explainability and a easy solution to train and proper the mannequin in a clear consumer interface. From constructing on prime of reliable basis fashions on the base proper by to easy-to-use testing and measurement instruments, SeekrFlow ensures correct outcomes that may be understood and validated. It’s important to know that AI guardrails aren’t simply one thing that’s good to have or to consider later – relatively, we provide prospects easy to make use of explainability and contestability instruments from the beginning of implementation.

How does the platform combine information preparation, fine-tuning, internet hosting, and inference to allow quicker experimentation and adaptation of LLMs?

SeekrFlow integrates the end-to-end AI growth course of, in a single platform. From dealing with the labeling and formatting of the information with its AI agent assisted technology method, to fine-tuning a base mannequin, all the way in which to serving for inference and monitoring the fine-tuned mannequin.  As well as, SeekrFlow’s explainability tooling permits AI modelers to find gaps within the data of the mannequin, perceive why errors and hallucinations happen, and immediately act upon them. This built-in, end-to-end method permits speedy experimentation and mannequin iterations

What different distinctive applied sciences or methodologies has Seekr developed to make sure the accuracy and reliability of its AI fashions?

Seekr has developed patented AI know-how for assessing the standard, reliability, bias, toxicity and veracity of content material, whether or not that could be a textual content, a visible, or an audio sign. This know-how supplies wealthy information and data that may be fused into any AI mannequin, within the type of coaching information, algorithms, or mannequin guardrails. In the end, Seekr’s know-how for assessing content material could be leveraged to make sure the security, factuality, helpfulness, unbiasedness and equity of AI fashions.  An instance of that is SeekrAlign, which helps manufacturers and publishers develop attain and income with accountable AI that appears on the context of content material by our patented Civility Scoring.

Seekr’s method to explainability ensures that AI mannequin responses are comprehensible and traceable. As AI fashions turn into concerned in choices of penalties, the necessity for AI modelers to know and contest mannequin choices, turns into more and more vital.

How does SeekrFlow’s precept alignment agent assist builders align AI fashions with their enterprise’s values and trade rules?

SeekrFlow’s precept alignment agent is a crucial characteristic that helps builders and enterprises scale back the general value of their RAG-based programs and effectively align their AI to their very own distinctive ideas, values, and trade rules with no need to assemble and course of structured information.

The Seekr agent makes use of superior alignment algorithms to make sure that LLMs’ conduct adheres to those distinctive and predefined requirements, intentions, guidelines, or values. In the course of the coaching and inference phases, the precept alignment agent guides customers by the complete information preparation and fine-tuning course of whereas repeatedly integrating professional enter and moral tips. This ensures that our AI fashions function inside acceptable boundaries.

By offering instruments to customise and implement these ideas, SeekrFlow empowers enterprises to keep up management over their AI functions, guaranteeing that they mirror the corporate’s values and cling to authorized and trade necessities. This functionality is crucial for constructing belief with prospects and stakeholders, because it demonstrates a dedication to accountable AI.

Are you able to focus on the collaboration with OneValley and the way Seekr’s know-how powers the Haystack AI platform?

OneValley is a trusted useful resource for tens of hundreds of entrepreneurs and small to medium sized companies (SMBs) worldwide. A typical drawback these leaders face is discovering the suitable recommendation, services and products to begin and develop their enterprise.  Seekr’s trade particular LLMs energy OneValley’s newest product Haystack AI, which gives prospects entry to huge databases of obtainable merchandise, their attributes, pricing, professionals and cons and extra. Haystack AI intelligently makes suggestions to customers and solutions questions, all accessible by an in-app chatbot.

What particular advantages does Haystack provide to startups and SMBs, and the way does Seekr’s AI improve these advantages?

Whether or not a consumer wants a quick reply to know which enterprise bank card gives the best money rewards with the bottom charges per consumer and lowest APR or to distinction and examine two CRM programs they’re contemplating as the suitable answer, Haystack AI powered by Seekr supplies them the suitable solutions shortly.

Haystack AI solutions customers’ questions quickly and in essentially the most cost-effective method. Having to attend for and ask a human to reply these questions and all of the analysis that goes into this kind of course of is unmanageable for terribly busy enterprise leaders. Clients need correct solutions they will depend on quick, with out having to trawl by the outcomes (and sponsored hyperlinks!) of an online search engine. Their time is greatest spent operating their core enterprise. It is a nice instance the place Seekr AI solves an actual enterprise want.

How does Seekr be certain that its AI options stay scalable and cost-effective for companies of all sizes?

The straightforward reply is to make sure scale and low value, you want a strategic collaboration for entry to compute at scale. Delivering scalable, cost-effective and dependable AI requires the wedding of best-in-class AI software program operating on main technology {hardware}. Our collaboration with Intel includes a multi-year dedication for entry to an ever-growing quantity of AI {hardware}, together with upgrading by generations from present Gaudi 2 to Gaudi 3 in early 2025 and onwards onto the following chip improvements. We positioned a wager that the perfect availability and worth of compute would come from the precise producer of the silicon, of which Intel is just one of two on the planet that produces its personal chips. This solves any points round shortage, particularly as we and our prospects scale and guarantee the very best worth efficiency that advantages the shopper.

Seekr prospects operating on their very own hosted service solely pay for precise utilization. We don’t cost for GPUs sat idle. SeekrFlow has a extremely aggressive pricing mannequin in comparison with contemporaries within the area, that helps the smallest to largest deployments.

Thanks for the nice interview, readers who want to study extra ought to go to Seekr.