Tower — a Telegram medium for students in architecture and design programs, young professionals, and emerging talent. It helps users navigate the job market and select architectural and design studios based not only on open vacancies but also on the nature of projects, visual language, specialization, work environment, and professional culture.
The core idea of the project is to transform the job search into a conscious choice of a professional environment. The media platform doesn’t just show where a position is open; it also details what the studio does, what projects it is executing, what skills are in demand, and how closely its direction aligns with the candidate’s interests.
->
Data sources for media
1/ Career Platforms. HeadHunter is one of the primary sources. Here, you can find relevant job openings, study employer requirements, experience levels, offered conditions, employment formats, and salary offers. HeadHunter’s open API provides an additional capability, allowing structured information about vacancies and companies to be retrieved for subsequent processing.
2/ Professional Telegram Channels. Specialized communities play a vital role in publishing job openings for architects and designers. Among these are WASA Jobs, «Kaltura-Architecture, ” and „ArchHunter.“ These sources allow users to find opportunities that might not be available on major career platforms, including internships, project work, and entry-level positions.
3/ Official websites of architectural bureaus. Employers' websites are used to obtain the most reliable information regarding specialization, team, completed projects, geographical scope of work, and open positions. This is the primary source that allows you to verify information found on third-party platforms and supplement it with details about the firm itself.
4/ Portfolio and Visual Platforms. Behance, ArchDaily, and other professional platforms are used to study the visual language of firms and analyze their projects. This data helps define the primary focus areas of the work, the nature of the architectural solutions, the typology of objects, and the characteristics of the design approach.
5/ Social Media and Professional Communities. Public pages of design studios and their staff serve as additional sources. These platforms allow users to track news, announcements about current projects, team expansions, contest participation, and professional events. Information gathered from these sources is considered supplemental and requires verification.
Data schema divided into human and machine processing
Phase 1. Information Gathering. In the first phase, the system collects data from career platforms, professional Telegram channels, architectural firm websites, and visual hubs. Machine processing is used to search for new publications and extract text, links, images, and key job characteristics. Human curators can additionally identify interesting firms and materials that automated search misses.
Phase 2. Data Extraction and Structuring. Automated tools extract the firm’s name, job title, candidate requirements, work experience, specialization, employment type, salary, and contact information from the source materials. This gathered information is then organized into a unified database. An editor verifies the accuracy of the extracted data and completes fields that were missing from the original publication.
Phase 3. Analysis and Classification. AI assists in categorizing job openings by discipline: architecture, interior design, landscape architecture, urbanism, environmental design, and other specializations. Additionally, the system can determine the seniority level of the position, required software, project type, and suitability of the opening for entry-level specialists. A human reviews the classification results because job titles and employer requirements are not always straightforward.
Phase 4. Architectural Firm Profiling. The system links job openings to employer profiles and compiles information about their projects, specialization, and visual language. Machine tools can analyze descriptions and classify materials, but the editor provides the final interpretation of the firm’s design approach. This is especially—
General Data Flow Diagram
Information Sources → Automatic Collection → Data Extraction → Database Structuring → AI Classification → Editorial Review → Publication Preparation → Telegram Publication → Audience Reaction Analysis.
Guidelines for Automated Services in the Scheme
4.1. ChatGPT — Text Analysis and Preparation
ChatGPT is used to process gathered information, analyze job postings, prepare agency profiles, compare employers, and draft publications. The service receives a job posting, company description, project links, and other verified materials as input. The output must be a structured text that includes a brief description of the position, requirements, working conditions, and agency characteristics.
Guideline: The service must use only the provided sources and must not add missing facts. If salary, schedule, or working conditions are not specified, this must be stated explicitly. The tone of the publications should be professional, accessible, and informal—avoiding bureaucratic jargon and advertising hyperbole. All text is reviewed by an editor before publication.
4.2. Parsers and APIs — Data Collection
Open interfaces and data gathering tools are used for automated data retrieval. For instance, the HeadHunter API allows for structured information acquisition regarding vacancies and employers. For other sources, available export mechanisms, RSS feeds, notifications, and manual submission of publications into the system are employed.
Guideline: The service must collect only the information necessary for media processing: company name, position, requirements, terms, publication date, and a link to the original source. The original link and the date the data was collected must be preserved. Recurring job postings must be identified and consolidated, while outdated listings must be flagged for re-verification. Data collection must account for access restrictions and the usage policies of third-party platforms.
4.3. OCR and Image Recognition — Text Extraction
Parts of vacancies and announcements are published as images, presentations, and graphic cards. OCR tools are used to extract information from such materials, allowing the text within images to be converted into an editable format.
Guideline: The service must recognize the text, preserve the original information sequence, and highlight the main job posting fields. If the text is not fully recognized or contains errors, the result must be flagged for manual review. The OCR should not interpret missing data or reconstruct unknown job conditions independently.
4.4. Google Sheets or Airtable — Data Storage and Structuring
A spreadsheet or a specialized database is used to organize the database. It stores information about job openings, architectural firms, projects, and information sources. Each entry receives a unique identifier, an update date, and a link to the original source.
Guideline: data must be stored in a unified format. For job openings, the fields used are «agency, ” „position, ” „specialization, ” „experience, ” „salary, ” „work format, ” „requirements, ” „publication date, ” and „link.“ For the agency profile, the fields are „name, ” „areas of operation, ” „key projects, ” „visual language, ” „website, ” and „open vacancies.“ Changes must be tracked so the editor can monitor the currency of the information.
4.5. Telegram Bot API — Publication Automation
Telegram is used as the primary platform for content distribution. The Telegram Bot API can be employed for automating publications. It allows for sending prepared messages, images, and links to the channel, as well as organizing interaction with other system elements.
Guideline: The bot only receives materials that have passed editorial review. Publications must adhere to the unified toWer visual style, include correct links, and have a clear structure. Automatic sending of unverified job postings is prohibited. Before publishing, the system must check for required fields and the absence of duplicate materials.
4.6. Analytical Tools — Studying the Audience
To assess content effectiveness, we utilize Telegram’s built-in statistics and various analytics tools. These help determine which publication types generate the most interest: job openings, studio reviews, employer comparisons, or professional development materials.
Guideline: the analytical system must track views, reactions, shares, and link clicks, provided that the corresponding data is available. Recommendations for developing the content plan are generated based on this statistics. However, metrics should not become the sole criterion for quality—the editorial team also assesses the usefulness and accuracy of the materials.
Final Project Logic
toWer uses artificial intelligence not as an independent media author, but as a tool to organize professional information. Automation aids in finding job openings, extracting data, classifying offers, and preparing materials. The human remains responsible for fact-checking, analyzing architectural designs, editorial decisions, and final publication.
Thus, toWer unifies several disparate streams of information into a single system. The user receives not merely a list of vacancies, but a structured overview of employers, their projects, requirements, and professional environment.
