AI agent for collecting and structuring information

Hello! My name is Nikita. I have been implementing AI solutions in paid advertising and automating marketing processes for over 2 years, working with Google Ads, Meta Ads, and TikTok Ads.
✅ What you get when working with me:
— AI-enhanced advertising strategy instead of chaotic launches
— automation of analytics and control of project economics
— systematic scaling based on data and AI tools
📈 I work with projects of various scales and use AI for faster analysis of results, finding growth points, and optimizing advertising processes without unnecessary costs.
I am ready to discuss your tasks and offer a practical plan for implementing AI in your project's advertising.

Show original

Translate

Hello! A system that automatically scans a dozen sources, catches new information, and delivers only what is truly worth attention, with a brief note on why — this is exactly what I have been working on.

The main focus here is not on parsing (which everyone can do), but on the logic of selection: what is signal, what is noise, and what requires action. I have already implemented this in a live project called ai-radar — it’s a bot on RAG with hybrid search that collects sources, removes duplicates and irrelevant content, and provides a concise output with conclusions. You can check it out here: auth_ai_radar_bot (tg)

Here’s how I envision your MVP:

- A planner that regularly scans your 10+ sources;
- Detection of new information and changes on pages, without repeating what has already been seen;
- AI layer: a brief summary plus a category (news, product, partnership, job, tender, report, important signal) and a flag when something requires action;
… - Recording in Google Sheets, Airtable, or Notion — whichever is more convenient for you;
- A short digest in Telegram or via email;
- Instructions on how to add new sources independently.
Stack: Python with LLM API (Claude or GPT) as the core of selection, n8n will be connected where it truly simplifies things.

GitHub: DmytroVrd ·

Show original

Translate

Creating automated parsing and intelligent content filtering systems is a proven solution for cutting through information noise and highlighting critical business signals. I have significant experience in developing custom integration systems and data exchange architecture, including setting up complex multi-stage automation scenarios on n8n and Make platforms. For your MVP, I will build a stable data collection logic, where a key stage will be algorithmic scoring and strict classification of each finding before recording it in the database. Please let me know which 10 sources are prioritized in the first stage, and whether any of them have complex dynamic content or protection against automated reading. All filtered signals will be clearly structured in Google Sheets by types (tenders, job vacancies, products), and a concise summary digest will be instantly sent to your email. The implementation of a working MVP along with the preparation of instructions will take 4 working days, and the cost of the work is 9,000 UAH. Let's discuss the technical details of the first sources to launch monitoring soon.

Show original

Translate

I will create a smart MVP for monitoring based on n8n, which will actively filter out the "noise," delivering only high-value, categorized signals from competitors directly to your Telegram.

Instead of drowning in raw scraped data, your team will receive actionable insights with brief AI summaries (tenders, product updates, partnerships). This eliminates hours of manual checks and ensures that you never miss a strategic move in your industry.

How I will implement the MVP:
To avoid the standard trap of "scraped → recorded," I will build a highly logical pipeline using n8n as the central orchestrator:

Target data collection: I will set up scheduled checks for 10+ sources. Depending on the complexity, I will use native HTTP requests in n8n or specialized tools to detect actual content changes.

… AI signal filtering and classification: This is the core. The detected text will be processed by OpenAI/Gemini with strict system instructions. The AI will analyze the context, discard the "noise" (e.g., minor interface changes, general PR fluff), make a concise conclusion about the real value, and classify it into the categories you need (news, product, job vacancy, etc.).

Storage and notifications: The structured data will be stored in a clean Airtable database. Finally, a Telegram bot will send a formatted, easy-to-read digest.

Answers to your specific questions:
Have I created similar systems? Yes. As a certified n8n automation expert, building custom AI pipelines for data extraction and enrichment is my main focus.

Case example: Recently, I developed an automated scraping and recording pipeline into a RAG database, which extracts unstructured data from website text, uses LLM to normalize it and filter out irrelevant information, and records the structured result in the database.

Proposed implementation: As detailed above (n8n + AI classification prompts + Airtable + Telegram). I will also provide clear instructions on how you can add new URLs to the n8n process yourself.

Estimated budget and timeline: For a quality, fully functional MVP covering 10 sources, I estimate the work to be around $100-300 depending on the website protections and 10-14 days of development.

Do any of the initial 10 sources have strict bot protection (e.g., Cloudflare)? Let’s have a brief chat to agree on the exact criteria of what you consider a "useful signal" so we can properly launch the MVP.

Best regards,
Ivan

Show original

Translate

Hello! I can fulfill your order as I have practical experience in creating autonomous AI agents, data monitoring systems (Scraping/Parsing), and building automated funnels based on n8n and OpenAI. Previously, I implemented a similar system for an e-commerce project (AI assistant on n8n + OpenAI), where I set up regular synchronization, filtering, and processing of large volumes of text data. The principle of filtering out informational noise using LLM is completely familiar to me.

Logic of the script:

Data collection (No-code Parsing): For sites with RSS feeds, we use the native n8n node (this is free and instant). For complex sites without RSS or competitor pages, we connect Apify or a lightweight custom HTTP request in n8n (extracting HTML text via HTML-to-Text).

Intelligent AI filtering (Anti-noise): Instead of dumping everything indiscriminately, we pass the text to GPT-4o mini with a strict system prompt and structured output (Structured Outputs / JSON). The model, in one request:

Evaluates importance (0 — spam/marketing noise, 1 — important signal). If 0 — the process stops.
…
Classifies according to your tags (job, tender, news, etc.).

Creates a brief summary (TL;DR) in 2-3 sentences.

Storage and Notification: Filtered useful signals are recorded in Google Sheets (or Airtable), and a beautiful notification with tags and a link is sent to the Telegram channel.

I will specify the budget and deadlines in the feedback, my GitHub: https://github.com/onyx144

Show original

Translate

Good day.

We can implement such an MVP. This is a task that is quite close to us: not just gathering information from websites, but filtering out the noise, structuring the signals, and delivering only what is truly worth attention.

We have a relevant internal case: we are currently developing our own Telegram tool for monitoring new freelance projects. It receives new entries, removes duplicates, classifies them by relevance, allows searching by categories, stores statuses, notes, proposal texts, and prepares Telegram notifications for important signals. The logic is very similar: sources → new data → AI/rules → signal or noise → table/bot/action.

Here’s how I see the MVP for your task:

1. Sources
… We connect 10+ websites or pages. For simple sources, we can use RSS/HTTP/n8n, for more complex ones — Apify/Browse.ai or a lightweight parser.

2. New Detection
The system checks the pages on a schedule, records new publications, documents, or changes, and does not duplicate already found entries.

3. AI Analysis
OpenAI/Claude analyzes the found material, makes a brief summary, and determines the category: news, product, partnership, vacancy, tender, report, important signal, or noise.

4. Structured Database
The result is recorded in Google Sheets / Airtable / Notion: source, date, link, category, brief summary, importance, whether action is needed.

5. Telegram/email Digest
A short digest is sent to Telegram or email only for new and useful signals, not for every minor change.

For the first version, I would create a simple and maintainable architecture: n8n or a Python script as the orchestrator, Google Sheets as the database, OpenAI for analysis, Telegram Bot for notifications. After the MVP, we can expand the filtering logic, add more sources, separate categories, priorities, and daily digests.

We are approximately ready to create the MVP with 10 sources for 4,000 UAH / 5 days, provided the sources do not have complex authorization and aggressive anti-bot protection.

To start, we need to clarify:

- which specific 10 sources we will take for the MVP;
- how often to check them;
- where it is more convenient to write the results: Google Sheets, Airtable, or Notion;
- whether the digest is needed in Telegram or email;
- what examples are considered "important signals" for you, and what is considered noise.

Show original

Translate

Hello! This project is exactly what I specialize in. Building smart automated connections using AI is my main profile.

1. Have I created similar systems and an example case:
Yes, I have extensive experience working with Make.com, APIs of various neural networks (OpenAI, Claude, etc.), as well as Google Sheets and Telegram. Most recently, I developed a complex automated assistant for a business (service sector). The system had a complex routing logic, processed incoming data via API, interacted with databases in Google Sheets, and automatically sent notifications in Telegram.

2. How I propose to implement your MVP:

Orchestrator: Make.com. This is a reliable platform for building complex scenarios.

… Data collection (10 sources): Depending on the structure of the websites, we will use built-in Make modules (HTTP/RSS) or connect lightweight scrapers like Apify if the sites are dynamic.

Analysis and filtering (Core of the system): Data is sent via API to OpenAI. I will write a detailed system prompt that will make the AI act as an analyst: filtering out noise (for example, footer updates or privacy policy changes) and highlighting only target signals.

Categorization and Summary: The same prompt will make the AI determine the type (tender, vacancy, news) and generate a concise summary in 1-2 sentences.

Database and Notifications: The filtered result is recorded in Google Sheets for history, and the most important signals are instantly (or once a day) sent via a Telegram bot in the form of a convenient digest.

After completing the MVP, I will prepare a clear video instruction or text guide on how you can independently add new sites for monitoring in Make.com.

3. Timeline and budget:
To build and test the MVP for 10 sources, I will need approximately 3-5 working days. The exact budget and final choice of parsing tool I will be able to provide after I look at the list of these 10 websites (to assess their technical complexity for data collection).

I am ready to discuss the details and start working. Write to me, and I will show you how we will set up the logic for selecting "signals."

Show original

Translate

Good day!

Yes, I have built such systems. One example is automated monitoring of 30+ external sources with a daily digest in Telegram: each source is checked on a schedule, new entries are deduplicated, AI briefly summarizes the essence and tags it (news / warning / action needed). The result is written in a table, and only what is truly new goes to Telegram.

Here’s how I see the MVP for you:
I recommend n8n — it’s easy to add new sites without coding, there are ready-made nodes for Google Sheets, Telegram, HTTP.
• Cron trigger → check each site (HTTP / RSS / Apify for complex pages)
• Comparison with the previous state → only new changes go further
• AI (OpenAI): brief summary + category (news / product / partnership / vacancy / tender / report / signal / noise)
… • Record in Google Sheets / Airtable
• Telegram digest once a day or instantly for "important signals"

Three clarifications that affect the approach:
1. What types of sites are on the list — are there RSS/Atom feeds, or are they mostly regular HTML pages without an API?
2. Where to run n8n — cloud (n8n.cloud) or self-hosted on your server?
3. Is there an OpenAI API key, or should its cost also be considered?

MVP (10 sources, AI classification, Google Sheets + Telegram, instructions for adding new sites) — 4,000 UAH, deadline 5–6 working days. I work through a Secure Agreement.

If the result is suitable — I would gladly continue: more sources, more complex prioritization logic, additional notification channels.

Show original

Translate

Most automations in this field fail due to the lack of a minimum signal selection system. For the MVP, it is necessary to combine three components: change recognition (Apify/Playwright), AI analysis (OpenAI/Perplexity for classification), and a logical noise filtering stage.

Implementation: We will set up Apify as a monitoring script, with results going to Google Sheets via n8n, supporting Airtable/Notion. A Telegram bot will sort the results by categories before delivery. The key part is configuring the AI model to distinguish news from tenders.

Issues at the MVP stage: the number of sources limits API quotas, a proxy pool is needed, as well as manual configuration for sites with dynamic content. In past projects, this took 60–80 hours.

Which sources are already being monitored regularly? It is important to know if there is access to them via API or if they need to be collected through a headless browser.

Show original

Translate

Good day, I have already worked with similar automated monitoring and signal selection systems: data collection from websites, checking for updates, noise filtering, event classification, and sending results to tables and messengers. My main background is in Python, API integrations, automation, AI analysis, and building working MVPs for regular source monitoring.

An example of a similar case without an NDA: I created systems where it was necessary to regularly fetch data from external sources, track new entities or changes, normalize them, apply prioritization logic, and pass the result to Telegram / a table / an internal interface for further work. I also have practical experience in projects where not just integration is important, but specifically the logic of selecting useful events and automating actions after detecting a signal.

Show original

Translate

Hello
write to me, I will do it
on n8n + Apify/OpenAI + Google Sheets + Telegram.
check the reviews, everything is always great

Show original

Translate

Hello, Pavlo! The task is completely clear. I have experience in building exactly such smart pipelines, where the key is not just parsing everything randomly, but specifically AI filtering of informational noise and highlighting triggers that require action.

Show original

Translate

Good day!

Yes, I have worked with similar AI automation systems where it is necessary not just to collect data, but to analyze it, filter important signals, and automatically transmit the results to a CRM or knowledge base.

A similar case is the development of an AI marketing system that integrated Make.com, OpenAI, CRM, and automated data processing scripts. I also implemented a CRM ecosystem with automatic lead collection, AI processing, tagging, and analytics without manual intervention.

I would implement the MVP on Make.com or n8n using Apify/Browse.ai for monitoring sources, OpenAI for analyzing and classifying signals, Google Sheets or Airtable for storing results, and a Telegram Bot for instant digests. Additionally, I would add AI filtering to cut out informational noise and leave only the events that require attention.

The estimated development time for the MVP is 5–7 days, with a budget of $400–600 depending on the number of sources and the required analysis logic.
…
I would be happy to discuss the requirements and propose an architecture that is easy to scale after the successful launch of the MVP.

Show original

Translate

Hello! I can implement an MVP for a website monitoring system using n8n or Make.

I have experience in AI automation using n8n, Make, OpenAI API, Telegram, and Google Sheets. One of the relevant cases is AI processing with data analysis and structuring for CRM: https://freelancehunt.com/showcase/work/make-integromat-ai-obrobka-pdf-ta/2044836.html

Please let me know which sources you plan to connect in the MVP so I can suggest an architecture.

Show original

Translate

Good day!

I have built such systems — specifically with signal selection logic, not just "scraped → recorded".

**Similar case:**
I did monitoring for a company that tracked 15+ competitor websites and industry portals. The system checked new publications daily, compared changes on pages using diff logic, sent text to OpenAI for classification and a brief summary, filtered out noise (insignificant layout changes, date updates), and sent a digest to Telegram with only truly important signals. Results were written in Google Sheets with categories and priority.

**How I will implement your MVP:**
Stack: Python + OpenAI API + Google Sheets API + Telegram Bot.
…
— The list of websites is stored in Google Sheets — it’s easy to add new sources without touching the code
— The parser checks each site on a schedule (cron), records new publications and changes by comparing with the previous state
— OpenAI analyzes each signal: brief summary + classification (news / product / partnership / vacancy / tender / report / important signal)
— Additional filtering logic: if the change is insignificant or repetitive — we don’t write, we don’t create noise
— The result is written to a table, once a day/week (as you wish) — digest in Telegram

**Budget and timeline:**
MVP with 10 sources — 7 working days, $300–400.
Includes: working code, results table, instructions on how to add websites, configured Telegram bot.

If the result is satisfactory — I am ready for long-term work on the system.

I am ready to answer questions or show an example of the results table structure.

Show original

Translate

Good day.

Your project is close to my stack and implementation logic. I work with automations, AI integrations, n8n, APIs, Telegram bots, Google Sheets, and scripts for gathering and structuring information from various sources.

I can implement an MVP of such an agent:
- regular checking of specified sources;
- detection of new publications/changes;
- brief AI analysis of the found content;
- basic classification by types;
… - recording results in Google Sheets/Airtable/Notion;
- sending a brief digest via Telegram or email.

I see the first stage optimally as a working MVP to quickly test the logic on real sources and only after that scale the system.

To correctly assess the scope, I want to clarify:
1. How many sources need to be monitored at the start?
2. What specific types of sources: websites, blogs, news pages, PDFs, documents?
3. What classification is mandatory at the first stage?
4. Where should the results be prioritized for recording: Google Sheets, Airtable, or Notion?
5. Where should the digest be sent: Telegram or email?

I can take on the first working version and build it in such a way that it will be easy to expand the number of sources, types of classification, and reporting formats later.

I am ready to discuss the details and propose the simplest and most reliable implementation option within your budget and goals.

Show original

Translate

Hello! I can build such an MVP.

I have experience with similar logic: regular monitoring of sources, detecting new publications/changes, deduplication, brief AI analysis, and recording results in a table with a message in Telegram.

Here’s how I see the implementation of the MVP:
1. A list of 10+ sources and a check planner.
2. Gathering new pages/changes via Python or n8n/Apify, depending on the websites.
3. Deduplication to avoid sending old or duplicate signals.
4. AI layer: brief summary, category (news / product / partnership / vacancy / tender / report / important signal), importance rating, and a "requires action" tag.
… 5. Recording in Google Sheets / Airtable / Notion and a brief digest in Telegram or email.
6. Instructions on how to add new sources.

To start, I would suggest creating a simple but functional version with 10 sources, after which we can enhance the rules, filters, and classification quality. Approximately 5 days, budget 4000 UAH.

Show original

Translate

Hello
Please provide 1-2 links to the websites you need
In response, I will give an example of the result obtained
I suggest doing it on n8n
Cost and deadlines in private messages

Show original

Translate

Good day. I am ready to implement.
I have already completed similar tasks:
- https://gloap.net/news/ - all news and images are created and published automatically based on data from other websites
- https://o-keto.com/news/ - all TOP-5 studies and images are generated and published automatically based on data from two other websites
- https://o-keto.com/ - AI-nutritionist from RAG database based on Google Docs
- https://gloap.net/ - AI-job search and AI-resume selection from RAG database
- https://gloap.net/ - AI-recruiter: searching for sailors on external websites and communicating in messengers
- vraki.net - over 100 different parsers, 95% of the website's content is parsed
- obuvnov.ru - a system with over 100 million products, updated daily through feeds
…
I can assemble an MVP turnkey: source scraping, change detection, data structuring, aggregation, and convenient result delivery. If there is a detailed list of sources and a scenario for the result, I will be able to estimate the timeline and cost more accurately.

Show original

Translate

Good day. I have worked on similar scenarios using n8n: data collection from sources, AI content analysis, selection of useful signals, and creating digests in Telegram and tables.

For this MVP, I would suggest building a flow: monitoring changes → AI noise filtering → categorization → recording results → automatic notifications.

I can quickly connect and start implementation today.

Show original

Translate

Hello!

The task you described is right in my profile. I have about 3 years of experience in developing automation systems and AI monitoring, so I understand perfectly how to set up the selection logic so that the neural network filters out 90% of the noise and captures only real signals.

Recently, I implemented a similar case: an AI scoring system for websites and RSS feeds that analyzed content based on triggers, categorized it, and sent ready notifications.

Here’s what I propose:
We will build an MVP based on n8n and OpenAI/Claude API/or other options within the budget. We will connect the monitoring of the necessary websites, AI analysis with a clear categorization, auto-recording in Google Sheets, and instant structured notifications to your Telegram bot. After launch, I will prepare a simple guide so you can easily add new sites yourself.

… Conditions:
Timeline: 3–5 days to launch a working MVP turn-key.

Budget: I am ready to adjust to your capabilities; let’s discuss the final price in private messages based on your budget and the list of websites.

Security and results: we can do the project through a safe.

Send the list of sources in private messages — we will discuss the details and start right away!

Show original

Translate

Good day, Pavlo! I have already gathered a similar monitoring system — it checked a list of websites and news feeds daily, caught new publications and changes on the pages, ran the text through OpenAI for a brief summary and tags (news, tender, vacancy, partnership, etc.), and compiled everything into Google Sheets with a morning digest in Telegram. The most important thing here, as you mentioned, is not the parsing itself, but filtering signal/noise — I do this through AI assessment of relevance based on your criteria plus deduplication, so only what is worth attention ends up in the digest, not every little detail. For the MVP, I would use n8n as the backbone (this makes it easiest to add new sources) plus OpenAI for analysis — 10 sources, a table with results, a Telegram bot, and a brief instruction on how to connect additional websites. Approximately within 5 days under your budget, and then we can calmly grow into ongoing improvements as desired. Just let me know — are the sources mostly regular websites and news feeds, or are there any closed ones that require a login?

Show original

Translate

Good day, Pavlo. I am interested in your project and would be happy to collaborate.
I have extensive experience in parsing/scraping content. I have worked with both news and closed resources.

1. I have created similar systems based on N8N. The full stack of tools will depend on the level of website protection and the amount of information needed.

2. I have experience collecting data from RSS, HTML, visual elements, searching for target URLs on closed sites, emulating users, and using proxies.

3. All processes within N8N.
The parsing tool will depend on the type of data needed - if nothing is closed, perplexity with clear limitations is sufficient. If something more serious and protected is required - ScrapingBee and similar tools.
… A script for the agent that will parse information - target topics, pages.
For analyzing the obtained data - gpt4.1-5.4 mini, depending on the type and volume.
Storage of processed data - Google Sheets for tests, scaling on Notion, creating a database and categorization.
Telegram bot - through N8N.

4. The budget and timelines depend on the type of data and the level of protection. The estimated cost starts from 5000 UAH.

I am ready to discuss the project details.
Have a great day!)

Show original

Translate

Good day!! I have experience integrating AI tools into work systems, and I have completed projects that can be viewed in my profile!! Feel free to reach out!!!

Show original

Translate

We already have a nearly ready similar solution for monitoring sources that can be quickly adapted and launched for your MVP. We can discuss the details here on the marketplace; I am available (:

The estimated MVP for 10 sources I would value at 18,000 UAH and 7 working days.

We have already created similar systems where it is necessary not just to collect pages but to separate useful signals from noise, briefly explain the essence, and deliver the result in a table or message.

For implementation, I see a simple first stage - n8n or Make for scheduling and integrations, Apify or a lightweight parser for more complex sites, OpenAI API for a brief summary, type of signal, and priority, then Google Sheets or Airtable and notifications in Telegram or email.

Look, here’s the nuance - for quality, we will need to agree on 5-7 examples of what is a significant signal for you and what is noise.
…
I would like to clarify 2 points.

- The sources will mainly be news pages, competitor pages, or PDF documents that also need to be monitored.
- Is only a table and a digest needed, or also statuses like requires action, verified, outdated?

Similar examples from Ingello.

- https://business.ingello.com/vorfahr - automation and AI logic for business processes, close to your task of selecting useful signals.
- https://business.ingello.com/fractal - agent processes and working with repetitive scenarios, similar in approach to AI classification.

The main page for the systems - https://systems-fl.ingello.com/ua

Show original

Translate

Good day! This is exactly the type of systems I have built — it is being implemented.
I will create the MVP using n8n + Apify + OpenAI API. The logic is as follows: regular source checks, AI analysis of each signal with classification by types, noise filtering, and only important information will be sent to Telegram and recorded in Google Sheets with a brief summary.
10 sources at the start is a reasonable volume for the MVP. After delivery, you will receive instructions on how to independently add new sites without a programmer. If the result is satisfactory, I am ready for long-term improvements.
Please clarify which specific sites need to be monitored and where it is more convenient to receive the digest — Telegram or email? I am ready to discuss the details.

Show original

Translate

Good day! We have experience in developing monitoring systems based on LLM and parsing tools. We implement this through Python agents using LangChain for data structuring and integration with APIs for automating news collection. We will build a reliable MVP that ensures accurate monitoring of your sources. We are ready to discuss the details of the tech stack for your project.

Show original

Translate

Good day, I am ready to complete your task quickly and efficiently. I have extensive experience in creating various bots. Please write to me in private messages to discuss the details. I would be happy to help :)

Show original

Translate