ITQuick: Jumse Automated Technical Screening Platform

The main articles are:

• Speech Recognition (Technology, Market)
• Speech technology: On the path from recognition to understanding

2025: System for analyzing more than 80 candidate behavior parameters during technical screening

On December 16, 2025, ITQuick (Aitikvik) announced the launch of the 2026 update of its Jumse platform, which implements multi-level protection against dishonest testing using neural networks and artificial intelligence models, as well as expanded tools for generating tasks and evaluating developers' competencies.

One of the key elements of the update was a system for analyzing more than 80 parameters of the candidate's behavior during technical screening, which allows you to accurately identify the use of third-party prompts and abnormal patterns of passing tests without using video surveillance.

To form an unlimited number of tasks, the platform uses AI agents built on the basis of its own methodology. The methodology was developed and tested by the expert council for three years. The content generation process consists of an average of 28 steps. It includes a detailed description of how, based on their own knowledge base (each of 22,000 thousand practical cases - tasks from the real practice of developers around the world), to form tasks with different formulations and answers. And then for each formulation, a multi-stage check is carried out for relevance and compliance with common sense.

The update also strengthened the module for building a detailed competency map: based on the results of technical screening, the platform forms a structured developer knowledge profile on the stack, topics and types of tasks, which simplifies decision-making for HR and technical directors. The platform is already used for automated determination of the level of developers with an accuracy of 98% based on the results of launches in large Russian technology companies.

As of December 2025, the Jumse platform supports more than 200 competencies, including popular programming languages, frameworks and infrastructure solutions. In 2026, the team plans to ensure technical screening in the main technologies and IT specialties presented on the Russian market, including in new areas related to artificial intelligence and data analysis.

Our task is to create a single, transparent and, most importantly, objective standard for assessing the technical competencies of developers for the market and make it resistant to using AI for dishonest testing, "said Oleg Lukavenko, director of technology at the Jumse platform.

2024: Implementing AI to create unique tests

On November 12, 2024, ITQuick announced a large-scale update to the Jumse automated technical screening platform, thanks to which the platform will allow creating unique tests for developers using artificial intelligence.

The technology solves the problem of uniformity of tests by speeding up the process of updating them and improving the accuracy of skill assessment. Now Jumse offers an unlimited number of test options, which makes the assessment of specialists more reliable and close to real work tasks.

Previously, tasks for candidates were formed manually, which limited the number of scenarios and slowed down the process of replenishing them. Jumse is built on a knowledge base containing more than 8,200 practice cases covering all key languages ​ ​ and technologies, including Python, Java and TypeScript. Each test was created with an emphasis on testing the skills required for the appropriate stack, understanding the key mechanisms and industrial specifics.

With the connection of artificial intelligence to the process of creating tests, we have taken an important step towards improving the accuracy of screening, "said Oleg Lukavenko, founder of Jumse. - Our goal is not only to evaluate the knowledge of developers, but also to see how they solve problems like those that they will face in real work.

The key element of the update is the development of a library of prompts that help AI create tasks based on the knowledge base and specifics of each task. Jumse's expert team has developed stack-specific and prompta-specific topics so that AI can generate tasks covering key aspects of programming and features important to various industries. This thoughtful structure allows you to vary not only the types of questions, but also the approaches to solving them.

The problem with working with AI is that without careful control it can issue answers that do not meet the real requirements of the task, - said Oleg Lukavenko. - Therefore, we entrusted the compilation and testing of promptas to our team, which is deeply versed in the context and needs of employers. The industrial library has become no less important for us than the cases themselves collected over three years of work.

The next step was to create an open interface that allows you to connect any language models (LLMs). This allows you to use the models that are most suitable for each specific technology. For example, Code Llama and Phind are used for Python, while Mistral and Wizard Code are used for Java and TypeScript. Such customization helps to improve the quality of the assessment, making it more adapted to the features of each language.