Tezi, an early-stage startup, is developing an AI agent to assist HR teams by screening resumes, scheduling interviews, and communicating with candidates
Today, the company disclosed a $9 million capital investment to facilitate its transition to a generally available product.
The alpha product is being introduced to a limited number of design clients this week; however, this is the company’s vision, as stated by CEO and co-founder Raghavendra Prabhu.
He acknowledged that HR has been utilizing automated resume screening for some time; however, Tezi recognized an opportunity to develop a more sophisticated recruiting tool for HR with the new iteration of large language models (LLMs).
“We believed that the combination of natural language and reasoning provided us with the opportunity to create something significantly different from the historical practices of software in this field,” he stated.
Jason James, his co-founder and COO, maintains that the current instruments are inadequate in his estimation. “Suppose you receive one thousand job applications.
He stated that in the past, AI, ML, or algorithms would have been capable of certifying that these portfolios are superb. “However, it is still necessary for a human to send emails, schedule interviews, and perform other tasks.”
In addition to fundamental ranking, it is now feasible to implement an end-to-end workflow.
The founders recognized that humans must remain involved in the process at this juncture, and they anticipate that it will become wholly automated as the models improve.
Additionally, the quality of the job descriptions and prompts will determine the pool of candidates that emerge from any job search.
Although they acknowledge that automation may result in bias, they are actively pursuing measures to mitigate this issue to the greatest extent feasible.
From their standpoint, they objectively evaluate the resumes based on the inputs provided by the recruiting manager. They acknowledge that they cannot regulate the appearance of the inputs; however, they assert that they are trying to reduce their own bias.
“At present, we are not well-equipped to prevent bias from the employer,” James stated. We will be taking measures to prevent the introduction of any bias into the equation through algorithmic means.
They are refraining from examining historical recruiting patterns. They intend for the models to align with the hiring manager’s criteria for talents and other attributes.
They have trained their models on 250 million profiles licensed by data providers. They have been working with OpenAI and Anthropic models and have tuned them to their hiring requirements.
The organization is currently in its infancy. It was introduced at the commencement of this year. They are commencing collaboration with 15-20 design clients with the expectation that they will resolve all issues and expand their beta distribution later this year.
Liquid 2, Afore, PrimeSet, South Park Commons, and industry investors contributed to the $9 million seed round led by 8VC and Audacious Ventures.