Марія Бровінська Startup 17 February 2025, 10:35

Ukrainian IT professionals have created a tool for analyzing resumes and interview transcripts with candidates. How Talently.ai makes the work of recruiters easier

Ukrainian developers have created an application for analyzing CVs, transcripts of interviews with candidates, and other related documents that are involved in the candidate selection process. The Talantly.ai application generates reports that highlight the candidate’s compliance with the requirements, his strengths and weaknesses. dev.ua talked to one of the Client Support Team Leads of the service, Ivan Pelypenko, and found out what the application is capable of and who can use it. Below is his direct speech.

The idea as a way to optimize hiring

Talantly.ai started with a practical need to optimize hiring. Initially, it was an internal initiative to solve our own recruitment challenges. We began experimenting with using AI linguistic models to more effectively assess candidates.

And then it became clear that not everything was that simple. So we expanded the range of analyzed data, adding interview transcripts from different stages of selection to resumes. And so we gradually developed «prompts» (a series of specific queries) for generating versatile reports on candidates.

The first version of the application was as simple as possible — a white screen with a few fields for uploading files and a space for displaying reports. This prototype, although not very user-friendly, was an important stage in the development. It was demonstrated to a small group of recruiters to illustrate the concept and get initial feedback on the idea and report formats.

The feedback we received prompted us to take a step back and start by creating a more presentable interface. Then we added basic functionality and improved reports. A month was allocated for this implementation, and right on time, an MVP (minimum viable product) appeared, ready for full testing.

Team

Talantly.ai is developed by Hitech Service, a company with 18 years of experience in the IT services sector. Despite its small size, the company has experience implementing projects of various scales — from small solutions to cooperation with such global giants as Kodak, ESKO and CBS. Initially, they were engaged in purely outsourcing (development, testing, administration and customer support), over time, the company gradually expanded its portfolio with its own products (for example, approval.studio).

Two opposing trends are clearly visible in the current recruiting tools market. The first is candidate-focused applications that offer automated resume creation, cover letter writing, and even interview preparation and near-interview execution. The second is fully automated systems that attempt to replace the recruiter by offering automated resume review, interviewing, and decision-making without human intervention.

We have chosen a fundamentally different approach. Instead of trying to replace the human recruiter, the product should become a tool to support professional recruiters and hiring managers. We are convinced that the experience of professionals in their field is invaluable.

The app should complement this experience by providing in-depth analysis of the candidate through reasoned metrics backed by real-world data from resumes, interviews, and other sources.

A feature of the product should be its ability to analyze information from various sources, revealing both potential «red flags» — alarm signals that may be unnoticed during a disparate analysis of documents, and the candidate’s hidden potential, which is revealed only with a comprehensive review of all available information.

Investment and time spent

The idea as it appears in the application and active development of talantly.ai began in late December 2024. After forming the concept and receiving initial feedback from recruiters, the team spent one month creating an MVP (minimum viable product). By the end of January 2025, the first full version of the product was created, ready for testing.

The project is implemented using internal resources of Hitech Service. Having a full development cycle within the company, various specialists were involved in the creation of talantly.ai. The flexible team structure allows us to effectively distribute resources according to the current needs of the project. Several people are involved in the project most of the time, but if necessary, we can afford to «pull out» the right specialist for a specific task.

Thanks to the use of its own infrastructure and resources, the main investment in the project is currently reduced to paying for tokens for the selected artificial intelligence model. It is worth noting that the team conducted thorough testing of various AI models and deliberately abandoned the use of OpenAI in favor of another solution that showed better results for the specific needs of the project.

MPV in action

At this stage, talantly.ai is in an active development phase. The development team is working on improving the MVP version of the product, focusing on three key areas: improving usability, ensuring stability, and expanding reporting functionality.

In parallel with the technical improvements, we made the app available to users and started inviting recruiters and hiring managers to join the product testing. This is an important milestone for the project, which has several goals:

Obtaining professional feedback to validate the product concept
Testing the system under real load
Collection of statistical data on resource use
Formation of a pool of wishes for additional functionality

The process of joining the testing is simple. Interested specialists can leave a request on the talantly.ai website in the Early Access section (https://talantly.ai/#access) or contact us directly via email ([email protected]) or LinkedIn (https://www.linkedin.com/in/john-ivan-pelypenko-78629896/). After that, we invite you to a short online meeting (10-20 minutes), during which the functionality of the application is demonstrated and the current limitations of the MVP version are explained. Interestingly, such demonstrations often develop into lively discussions, which inspires and strengthens faith in the relevance of the product for the industry.

Coming soon — release

At this stage, we are focused on incremental product development, without a hard-and-fast release date. This is a prudent approach, considering that the project is in its early stages of development — only a month and a half since the start of active development.

Before the public release, the team sets itself several key tasks:

Releasing a series of updates based on feedback from testers
Collecting and analyzing data on real-world product usage to assess its viability
Accumulation of statistics on the use of AI tokens and their value to form an optimal monetization model

It is important to note that the application architecture already includes an API interface, which opens up opportunities for future integrations with other services — the need for this was already voiced by the first testers of the product.

How it works

The process of creating a vacancy is as follows:

Generation of requirements based on the requirements of the job description and additional requirements based on the description of responsibilities in the job.
Uploading candidate resumes and receiving an initial assessment of the candidate’s compliance with the requirements.
Report on the compliance of the resume with the requirements of the vacancy. In addition to the assessment, the system makes assumptions about skills that indirectly meet the requirements, but are promising. Output of clarifying questions on weak positions.
When uploading an additional interview transcript, letter of recommendation, motivation letter, etc., a more detailed report will be generated that will take into account information from all available sources, build an evaluative judgment on it, citing quotes from these documents and confirming these judgments and assessments. The more documents, the better and more detailed the analysis.

As a result of such analysis, the recruiter receives an updated score of the candidate’s suitability for the vacancy.

Monetization

At this stage, the talantly.ai team is working on various models for monetizing the product. Although the prepayment or payment for used tokens model is common in the field of AI products, we are looking for a more convenient and classic solution for users. We are considering implementing a traditional monthly subscription with limits on the number of vacancies or candidates for analysis.

At the same time, the possibility of creating a hybrid payment system that would take into account different user needs is not excluded. A feature of talantly.ai is the flexibility for customization built into the architecture. This allows us to already offer individual solutions for large teams — for example, recruiting companies/departments that have certain methods of entering and evaluating candidates, which can be implemented in the form of reports in talantly.ai.

What’s next?

Today, talantly.ai already has basic functionality, which includes:

Creating and managing vacancies.
Formation of requirements for candidates (both automatically based on the job description and manually).
Maintaining candidate cards.
Initial analysis of the resume with an assessment of compliance with the requirements of the vacancy.
Suggesting clarifying questions to the candidate based on identified weaknesses.
Comprehensive analysis of the candidate based on all available materials (interviews, letters of recommendation and motivation, etc.).

The immediate plans for product development are focused on two key areas. The first is to improve the basic functionality of working with vacancies and candidates. The second is to expand analytical capabilities through the introduction of new types of reports:

SWOT analysis of candidates.
Identifying potential «red flags.»
Non-conformance report.
Analysis of ambitions and motivation.
Assessing cultural and behavioral fit.

Special emphasis is placed on the versatility of the analysis — each new type of report adds a new perspective on the candidate, helping the recruiter form the most complete picture of the potential employee.

"I have an avalanche of fake candidates." IT recruiters complain about the prevalence of candidates who pretend to be someone they are not. What's going on?

“A recruiter is not an NPC, we also have a memory.” A Ukrainian recruiter told what typical mistakes recruiters make in their work

"Save your time and recruiters, I beg you!" The recruiter said that she received a 39-page resume from a developer candidate. IT specialists claim that this is a good CV and joke