Soham Mazumdar, co-founder of Rubrik, who left in 2023, has a new data startup called Widdai. The company offers AI data analytics that can provide business insights with structured, unstructured, and even “dirty” data.
Where and how it uses data, it is essentially the holy grail of enterprise business intelligence software, and why Coatue led the huge $23 million seed round. Madrona, GTM Capital, and The Anthology Fund also participated.
Rather than asking the data analytics team to run the report, business managers can drill into more detail by asking Wisdoai questions.
Mazumdar cites the example of a person in charge of income who wants to know, “How are you planning on closing a quarter?” Widhidai’s answers provide a list of pending transactions that your team should focus on, along with information about what each customer is behind, including a list of questions they are waiting for.
“In contrast to the process involving a lot of time with five individuals, including some analysts, you can literally get a CRO through our platform, through our platform, to this last level of detail,” Mazumdar told TechCrunch.
It is just an example of the type of question that Wisdom wants to answer.
Another early customer is an oil and gas company with thousands of workers using Wisdomai to ask questions about production using Wisdomai.
TechCrunch Events
Berkeley, California
|
June 5th
Book now
Obviously, they also offer all the business analytics tools already available, as well as natural language prompts powered by AI.
Widdiai stands out for the founder’s pedigree. Everything used to work with Mazumdar on Rubrik. But the platform superpower is, even the accuracy for messy data, Mazumdar says. You can find answers not only in unstructured data stored in files, but also in structured data such as databases.
Equally important is that Widsomai does not bring hallucinations.
Most companies pursue the accuracy of AI apps by focusing on real-time search technologies such as data used to train AI models, model size, rapid engineering, and possibly searched generation (RAG). Still, they still take the risk of a processed answer.
Widdia uses Genai in query formations, but not in writing an answer. “In the end, genai can cause hallucinations. What we do with genai is to write small, small programs, which can query these different systems,” says Mazumdar.
So, if the widsomai model hallucinates, it’s all you need to write a fake query that failed to retrieve the data. The data itself – the answer to the question – is not manufactured.
Widhidai claims Conocophillips, Cisco and Descope are early customers, and they have customers working on major cloud data storage services such as Snowflake, Google’s BigQuery, Amazon’s Redshift, Databricks and Postgres. By studying the query language via query logs and other sources, Mazumdar says, it can be trained on any data storage system.