Make

We are engaged in the selection, purchase, and delivery of cars from the USA, Korea, and China. Currently, most inquiries are handled manually by managers: responses in chats, car selection, searching for lots at auctions, calculating costs, and qualifying clients. It is necessary to create an AI agent that can take on a larger part of this work and interact with clients as closely as possible to the logic of a live manager. The main goals of the AI Obtain the phone number. Subscribe the client to the appropriate ManyChat automation. Transfer the lead to Bitrix24. Answer questions. Keep the client engaged in the dialogue. Logic of operation Currently, clients can: choose a car through the ManyChat catalog; write manually what car they are looking for; go from an advertisement to a specific car. In all these cases, the AI should support the dialogue, clarify the need, and assist with the selection. Required experience Experience with the following will be an advantage: AI agents (OpenAI/Claude/Gemini); ManyChat; Bitrix24; API integrations; business process automation.

I need to create an automated SEO system based on n8n that will collect data from various services, combine it, and send it to ChatGPT for analysis. It is necessary to connect Ahrefs Google Search Console Google Analytics 4 Google Ads The first stage is a minimum viable version Connect Google Search Console. Connect GA4. Connect Ahrefs, if my plan and API allow it. Store daily data. Send it to ChatGPT. Send one daily report to Telegram or email. Set up analysis for the site viarcanvas.com and the competitor picturehappy.lv. After verifying the functionality, add Google Ads, additional countries, competitors, and extended reports. What needs to be obtained from Ahrefs For the site viarcanvas.com and selected competitors: organic keywords; current positions; position changes; search volume of queries; estimated traffic; pages that receive organic traffic; new and lost keywords; new and lost backlinks; competitor pages that have increased traffic; new competitor keywords; Content Gap between our site and competitors. The list of competitors should be editable without changing the n8n script itself.What needs to be obtained from Google Search Console clicks; impressions; CTR; average position; queries; pages; countries; devices; changes in metrics by days and periods. It is important that the data can be analyzed separately for each language version: /lv/ /lt/ /ee/ /pl/ /de/ A separate analysis of specific pages is also needed, for example: https://viarcanvas.com/lv/new/canvasWhat needs to be obtained from Google Analytics 4 users; sessions; organic traffic; landing pages; purchases; leads; revenue; conversion rate; engagement; countries and devices. It is necessary to link SEO traffic not only to visits but also to leads, orders, and revenue.What needs to be obtained from Google Ads spending; clicks; impressions; CTR; CPC; conversions; cost per conversion; revenue; ROAS; search queries; data on campaigns and countries. Google Ads should be analyzed separately from organic SEO but included in the overall marketing report.What ChatGPT should analyze ChatGPT should automatically determine: which keywords have increased; which keywords have decreased; which pages have lost traffic; which pages have started to grow; where impressions have decreased; where CTR has decreased; where the average position has worsened; which pages are in positions 4–15 and have the potential to reach the TOP; which keywords have impressions but few clicks; which competitor pages are growing faster than ours; which queries competitors have that we do not; which articles should be written; which existing pages should be updated; where query cannibalization is possible; where internal linking is needed; which pages should be strengthened with external links; which changes require urgent attention. ChatGPT should not just list numbers. It should explain the likely reasons for changes and provide a specific action plan.

We are manufacturers and sellers of men's clothing. We are developing an internal AI tool to automate the creation of advertising photo and video creatives for Facebook/Instagram. The project is being created for internal use, not for sale. What needs to be doneDevelop a user-friendly Dashboard where the user creates a project, fills out the product card (description, photo, price, target audience, etc.), after which a chain of AI agents is automatically launched, resulting in ready-made advertising photos and videos. Under the hoodAI Agent 1 — Creative Strategist Analysis of the product and target audience. Generation of Angles. Generation of Hooks. Generation of Big Ideas. Generation of Offers.AI Agent 2 — Script Generator Creation of advertising scripts. Hook. Story. CTA. Voice Over. Breakdown of the script by scenes.AI Agent 3 — Creative Director Generation of prompts for AI photos. Generation of prompts for AI videos. Creation of prompts for each scene. Automatic sending of prompts to AI for obtaining ready-made images and videos.All agents must automatically pass results to each other. Interface Dashboard. Project creation. Product card. Uploading product photos. Generation status. Gallery of ready-made photos. Gallery of ready-made videos. Option for re-generation. Working with colorsIt is necessary to implement convenient work with variants of a single product.For example:Linen suit Black White Mocha Graphite BlueFunctionality: Uploading one color. Uploading several colors of one product at once. Ability to choose which colors participate in generation. Choice of generation mode: only one color; even distribution among all selected colors; mixing colors in one creative.For example:The user selects 100 creatives.If 5 colors are uploaded, the system automatically distributes the generation: Black — 20 White — 20 Mocha — 20 Graphite — 20 Blue — 20Or it can create a creative where different colors of the product are automatically shown in different scenes. Generation settingsThe user must be able to choose: the number of creatives (10 / 25 / 50 / 100 or their own value); generate only photos; generate only videos; generate photos and videos simultaneously. Section "Scripts"For each ready creative, it is necessary to save: the finished script; Hook; Story; CTA; Voice Over; scene breakdown; separate ready prompt for each scene.This is necessary so that the editor can manually use these prompts to generate individual frames, images, or video clips in any AI services. Technical partIt is necessary to: connect the OpenAI / Claude / Gemini API; connect the image generation API; connect the video generation API; implement backend with automatic data transfer between AI agents; make the architecture scalable so that in the future it will be easy to add new AI agents and new AI models without a complete redesign of the project. When responding, please send: examples of AI projects; proposed technology stack; estimated development timelines; cost of MVP.Important: if you have experience in developing multi-agent AI systems or AI services with integration of multiple APIs, please indicate this in your response.

Development of a high-load (Highload) system with an LLM model for an online service of multimodal clothing search by photo and text query simultaneously integrated into messengers through a personal agent-assistant.

Develop an AI agent that autonomously executes user scenarios in the Chrome browser, identifies discrepancies with expected behavior, and formats them as bug reports in [Google Docs / Notion / Jira / Markdown in the repository]. How it should work: 1. Launch: [manually by button / scheduled once a day / automatically on each stage deployment]. 2. The agent opens Chrome, logs into a test account, and executes scenarios from the list (we will provide the list — [N] items at the start, format is free: in plain human language). 3. At each step, it checks the result: whether the expected element appeared, whether the data is correct, if there are any errors in the console and failed network requests, and whether the layout is broken. 4. At the end of the run, it generates a report: executed scenarios, found defects, for each defect — reproduction steps, expected and actual behavior, screenshot, severity, link to video or trace of the run Who can take this on and possibly even from existing repos on GitHub? Who is ready to start?