Blue and white data points representing a data fabric.

Illumex is using GenAI to ease pain of getting good data into LLMs

by with No Comment Tech

Blue and white data points representing a data fabric.

Image Credits: Yuichiro Chino / Getty Images

By now we know how crucial it is to have quality data for use by large language models (LLMs), but getting data ready for the models has been an early challenge for companies, an opening that represents an opportunity for an enterprising entrepreneur.

Enter Illumex, a two-year-old Israeli startup from the former VP of AI at Sisense. The startup is using GenAI to put the data into a ready state for LLMs. Today the company announced a $13 million investment.

Inna Tokarev-Sela, founder and CEO of Illumex, says she recognized this data readiness problem years ago, and she started Illumex with the goal of making it easier for organizations to organize data in an automated way.

“We automatically associate the business logic of an organization, automatically mapping it to data, and we bring the relevant data to the questions which business users have,” Tokarev-Sela told TechCrunch.

The company is combining a number of technologies to achieve this, including generative AI, graph databases and relational databases, pulling all this information together into what Tokarev-Sela calls a data fabric, which companies can access to train LLMs and for other purposes.

She sees the company as part of a continuum of the AI workflow, not simply getting data ready for the models, but also building applications based on this data, and ultimately deploying them with the goal of making it easier for end users to interact with and get what they need from the data Illumex is providing for them.

“We connect to all the systems, like different business applications and different business data sources, like databases, warehouses, data lakes, CRM, support systems, SAP enterprise financial systems and so on. So that provides us coverage of all data description and all logic and workflow description tightly coupled automatically in this knowledge graph,” she said.

Illumex view of data with different nodes connected to one another and the ability to filter by data type.
Illumex view of data in a graph format showing different connections between data.
Image Credits: Illumex

She says that this enables customers to get an inventory of their entire data landscape. “We provide an understanding of the semantics of the whole data landscape, like semantifying the whole stack, if you will, and then deploying generative AI with a high degree of accuracy and precision and context awareness,” she said.

What’s more, Illumex integrates with existing enterprise communications tools like Teams, Slack and Google Meet, so they are not forcing users to figure out a new way to work inside Illumex. The startup has customers using the platform, including highly regulated companies like financial services and pharmaceuticals.

The company currently has 30 employees and will be hiring with the new influx of capital. As a woman founder, Tokarev-Sela said she is trying to create a parent-friendly company, and reports that they are female-majority. “When you actually see the demographics of the company, it’s very different from the slogans. At Illumex, we actually walk the walk,” she said.

The $13 million investment was led by Cardumen Capital, Amdocs Ventures and Samsung Ventures, with participation from ICI Fund, Jibe Ventures, Iron Nation Fund, Ginossar Ventures, Icon Fund, Today Ventures and unnamed industry angels.

Illumex is using GenAI to ease pain of getting good data into LLMs

by with No Comment Tech

Blue and white data points representing a data fabric.

Image Credits: Yuichiro Chino / Getty Images

By now we know how crucial it is to have quality data for use by large language models (LLMs), but getting data ready for the models has been an early challenge for companies, an opening that represents an opportunity for an enterprising entrepreneur.

Enter Illumex, a two-year-old Israeli startup from the former VP of AI at Sisense. The startup is using GenAI to put the data into a ready state for LLMs. Today the company announced a $13 million investment.

Inna Tokarev-Sela, founder and CEO of Illumex, says she recognized this data readiness problem years ago, and she started Illumex with the goal of making it easier for organizations to organize data in an automated way.

“We automatically associate the business logic of an organization, automatically mapping it to data, and we bring the relevant data to the questions which business users have,” Tokarev-Sela told TechCrunch.

The company is combining a number of technologies to achieve this, including generative AI, graph databases and relational databases, pulling all this information together into what Tokarev-Sela calls a data fabric, which companies can access to train LLMs and for other purposes.

She sees the company as part of a continuum of the AI workflow, not simply getting data ready for the models, but also building applications based on this data, and ultimately deploying them with the goal of making it easier for end users to interact with and get what they need from the data Illumex is providing for them.

“We connect to all the systems, like different business applications and different business data sources, like databases, warehouses, data lakes, CRM, support systems, SAP enterprise financial systems and so on. So that provides us coverage of all data description and all logic and workflow description tightly coupled automatically in this knowledge graph,” she said.

Illumex view of data with different nodes connected to one another and the ability to filter by data type.
Illumex view of data in a graph format showing different connections between data.
Image Credits: Illumex

She says that this enables customers to get an inventory of their entire data landscape. “We provide an understanding of the semantics of the whole data landscape, like semantifying the whole stack, if you will, and then deploying generative AI with a high degree of accuracy and precision and context awareness,” she said.

What’s more, Illumex integrates with existing enterprise communications tools like Teams, Slack and Google Meet, so they are not forcing users to figure out a new way to work inside Illumex. The startup has customers using the platform, including highly regulated companies like financial services and pharmaceuticals.

The company currently has 30 employees and will be hiring with the new influx of capital. As a woman founder, Tokarev-Sela said she is trying to create a parent-friendly company, and reports that they are female-majority. “When you actually see the demographics of the company, it’s very different from the slogans. At Illumex, we actually walk the walk,” she said.

The $13 million investment was led by Cardumen Capital, Amdocs Ventures and Samsung Ventures, with participation from ICI Fund, Jibe Ventures, Iron Nation Fund, Ginossar Ventures, Icon Fund, Today Ventures and unnamed industry angels.

Knock takes the pain out of building notification workflows

by with No Comment Tech

Image Credits: Knock

Notifications may seem like a solved problem. You’re probably getting more than you want already, after all. The two founders of Knock, Sam Seely and Chris Bell, argue that while a lot of companies have solved the “last-mile delivery problem,” there is more work to be done. While products like Twilio and SendGrid may offer developer-friendly APIs, the Knock founders believe that what is really needed is a more comprehensive solution that combines notification delivery with a comprehensive workflow engine and integrated observability tools.

The company, which launched in 2021, today announced a $12 million funding round led by Craft Ventures. At launch, the company also raised a previously undisclosed $6 million seed round led by Afore Capital. Preface Ventures, Worklife,  Expa Ventures, CoFound Partners, and Tokyo Black also invested in these rounds, as well as angel investors like Vercel co-founder and CEO Guillermo Rauch and Behance co-founder Scott Belsky.

Image Credits: Knock

“Today, if you’re an engineering team for any type of product — whether it’s SaaS or developer tools, or consumer products — there are generic services that used to be built in-house and now you can go to an API for,” Knock CEO Sam Seely said. “Now, all the best engineers who want to work on payments, they go work at Stripe; and all the best ones that want to work on search, go to Algolia. It felt like the notification infrastructure part of this was still just the thing that you had to build in-house.”

Seely and Bell told me that they went back to the drawing board to see what a modern notification system would look like and what the primitives would be that they needed to build. At the end of the day, notifications aren’t a differentiator for most products, but they are very much a necessity. So if a product like Knock can speed up the development workflow, that’s a win-win.

The real differentiator for Knock is that it doesn’t just provide the tooling to send notifications but also pulls in data from third-party tools that can then trigger the workflow logic a developer has specified for their specific use case (like translating a message for a global audience).

Image Credits: Knock

“We have an entire workflow engine — that’s really the core of the product,” Seely explained. “That’s where you’re defining when some trigger happens. We call the Knock API, run through this workflow, batch messages on this cadence, throttle them so users don’t get spammed, and then send this in-app message, send this email message.”

This workflow engine is accessible through a web-based user interface, but as the team stressed, all of this functionality is also available programmatically. “A big kind of focus for us is taking that workflow engine that drives cross-channel engagement, but then bringing it into the everyday developer workflow,” said Seely.

Over time, Knock plans to go deeper into the customer engagement space, too. The team argues that every time a new channel emerges, existing players in this space — like SAP’s Exact Target, for example — have a hard time catching up.

“Users are getting tired of the onslaught and wave of emails and push notifications,” Seely said. “It’s real native product experiences that drive value to users and help companies drive engagement and retention — and all the reasons you send notifications in the first place. To drive native in-app experiences, that’s where developer experience matters.” And that’s where Knock thinks it can have a major advantage over the incumbents in this market. Seely noted that while the company often sees competitors like Iterable and Customer.io that are often sold to marketers, the secret of that market is that these tools are often mostly used and maintained by engineers.

One interesting aspect of the Knock tech stack: It’s written in the Elixir language, which isn’t exactly mainstream. As it turns out, Bell has long been very active in this community and even runs an Elixir podcast. “When I think about fit, in terms of what we’re building and the language choice, there has been no better application in my mind for the use of Elixir,” he explained. “Where it shines is this highly concurrent fault-[tolerance] that it brings to the table. When I think about what we’re doing here, the foundation of Erlang is written for telephony systems, routing calls from one place to another.”

The company plans to deploy the new funding to expand its go-to-market efforts and, of course, grow its engineering team. Current customers include the likes of Vercel, Amplitude, Hiive and Betterworks.