AI Solutions Architect / Senior Python Engineer (CrewAI / LangGraph / B2B Automation)
Looking for AI / Python Engineer: Local Multi-Agent System + B2B Sales Automation in WhatsApp (CrewAI / LangGraph)
Project: Development of an autonomous local ecosystem of AI agents for automating marketing, parsing regulatory data, and the full cycle of B2B sales for specialized websites for dental clinics.
- Current geo-focus: Italy (Lazio/Latina).
- Scaling: EU (Poland), USA, Canada, Australia.
🛠 Technology Stack:
- Frameworks: CrewAI or LangGraph (priority for long-term memory).
- WhatsApp Automation: Meta WhatsApp API, Green-API, Twilio or Whapi.cloud (with State Management setup).
- Scraping / Search: Firecrawl, BeautifulSoup, Tavily API / Google API.
- LLM & DB: GPT-4o / Claude 3.5, local ChromaDB.
- SMM APIs: Meta Graph API (Instagram/Facebook), TikTok Posting API.
📋 Key Tasks for Agents:
- 🔥 WhatsApp Sales Closer (Automated Sales): The system imports a database of dental clinics without websites. The AI agent initiates contact with the client on WhatsApp, engages in a full dialogue in the client's language (Italy, USA, etc.), handles objections, presents the agency's services, directs them to our website to choose a service, and independently closes the client for a call with the owner, sending a notification about the lead's readiness.
- Accounting & Law: Modules must dynamically gather information in real-time through web searches from specialized legal and tax portals of the respective country (current taxes, specific medical requirements, GDPR, HIPAA).
- Marketing & Automated SMM (TikTok, Instagram, Facebook): The AI receives a topic ➔ creates warming content ➔ automatically publishes posts/Reels/Shorts through official APIs.
- Personal Growth: Coaching the owner based on the integration of Python scripts for numerological (Pythagorean Square) and astrological calculations in the context of LLM.
🎯 Main Requirement:
To build a seamless funnel: from importing a "cold" phone number from the database to leading the client to the final call. The system must be architecturally ready for quick configuration changes for country scaling. Data must be stored locally on the PC.
📩 For contact: Please write in private messages, adding examples of your projects on CrewAI/LangGraph or cases of automating cold B2B sales in WhatsApp and website parsing. A detailed technical specification will be provided to selected candidates.
-
40 days609 USD
40 days609 USD
Regarding the project. The task is clear, but we need to immediately separate the wet fantasies from the CrewAI manuals and the harsh reality of production under high load and scaling. Creating a "set of scripts" that will get banned on WhatsApp in two hours is a matter of one evening. Building a fault-tolerant ecosystem that won't crash due to API limits and will maintain the states of thousands of leads is a completely different story.
What I recommend and how it MUST be done:
ARCHITECTURE BASE. No CrewAI for production in its pure form; they are only good for demos to investors. Only LangGraph. Why? We need strict State Management, control of cycles, and determinism of the graph. Agents should not go into infinite inference and eat up your money on LLM.
WHATSAPP AND BANS. Cold sales on WA without warming up and strict policies mean death for numbers. We use a proper gateway (the same Whapi or the official Meta API, if the budget allows) + OVERLAY control. We implement Human-in-the-loop at early stages for critical actions; otherwise, the neural network will promise clients free treatment at your expense.
SCALING AND LOCALITY. If it needs to be local, we package everything in Docker Compose. Local ChromaDB is okay for a start, but for proper RAG under the laws of different countries (GDPR, HIPAA), the database will need to be correctly sectioned by metadata so that the context of Italy does not mix with Australia.
…
Regarding demos and cases. All combat systems operate in closed client environments under strict NDAs; I won't provide access to "poke around" as it's a commercial secret. However, during a call with screen sharing, I can easily show the architecture of working systems, the logic of graphs, state machines, and how data flows.
To start and evaluate, we need to clarify three points:
Is there already a database of dental clinics (what format) or are we also writing a parser from scratch for maps/catalogs?
How ready are the texts of offers and scripts that the agent should learn to sell?
What hardware is allocated for local launch (are we setting up small local networks for parsing or is all inference strictly on OpenAI/Claude via API)?
If reliable software with solid logic is needed, not a crooked hack - let's discuss in private messages, and I will send the technical specifications for approval.
-
20 days609 USD
801 3 0 20 days609 USD
Good day, Andriy, you have quite a large project. So I will respond in order, structuring my answers:
1. I have worked with both LangGraph and CrewAI, but mostly I implement projects through PydanticAI. I can use the stack that you prefer.
2. I have also repeatedly worked with WhatsApp automation. I know how to work with their official API and Twilio. I already have ready-made Python modules for integration.
3. There are many options for searching and scraping. Let's consider this point in more detail:
3.1) Search: Vertex AI Search - for obtaining links + AI functionality for queries. DuckDuckGo API can also be used (to avoid falling into an information bubble). Brave is also available.
3.2) Scraping: for non-dynamic sites with poor/medium protection, there is Scrapling. It handles this task well and works well in combination with AI. The same can be said for Crawl4AI. For dynamic sites with strong protection — either third-party paid solutions (there are many). Or cloakbrowser — it effectively hides browser fingerprints.
4. For memory, rag, etc. — ChromaDB is decent, you can also look at PGvector and, simply, qdrant. For embedding, you can check rankings by languages and automatically select the optimal third-party service or local model based on geography.
Regarding implementation:
… In principle, you have already described it in detail. I don't quite understand what relates to numerology and coaching, but I have done a project related to AI generation in the field of esotericism. The task is clear, we have various modules and groups of agents for the tasks. Each has its own memory, its own context. Each performs its own functionality. We do everything adaptively right away. We add the necessary prompts, instructions, skills, and so on. I will also add auto-tests with LLM-as-Judge for the tasks.
From similar experience:
I created an AI advisor based on Ukrainian legislation.
The last project involved yacht analysis.
Content generation and forecasts for an esoteric application.
I would be happy to work with you!
-
1 day226 USD
1 day226 USD
Hello!
We are dZENcode – a full-cycle digital solutions development company: from design and programming to integrations and post-release support. We take on projects from scratch and can also assist in refining existing solutions.
We can engage in the task and create an AI solution for this funnel.
Are you considering involving an external contractor or team for these tasks? What tasks and technologies need to be prioritized?
You can find detailed information about our services and rates on our website:
Freelancehunt
… Take a look – after that, we can discuss the details and agree on the next steps.
⚠️ After clarifying all the details, we will determine the scope, the suitable format of cooperation: task-based, outsourcing, or outstaffing, and the final cost.
Why projects with us are guaranteed to reach release:
💎 10+ years of providing IT services;
🔥 90+ in-house specialists;
🚀 250+ public reviews since 2015;
⚙️ We support the product under SLA after launch;
✅ We work under NDA and contracts with the company!
-
45 days609 USD
45 days609 USD
This is our profile area. Here it is important not to create a set of scripts, but to build a system with architecture, memory, queues, logs, limits, roles, and dialogue verification before automated actions.
For the budget - the starting phase is 240,000 UAH, 45 days. In this phase, I would lay down the architecture, local knowledge base, and long-term memory, import dental clinics, the first WhatsApp agent, lead status, manual approval of critical messages, basic search for regulatory data by country, and a control panel. A complete ecosystem with SMM, multiple countries, a legal block, and automatic publications is better to do in the next phase - otherwise, the system will turn out fragile.
Main clarifications
> Which WhatsApp channel are you ready to use - Meta WhatsApp API, Twilio, Whapi.cloud, or another provider
> Is there already a database of dental clinics, an agency website, an offer, and examples of dialogues with leads
From your side, access to the API, a test WhatsApp number, the format of the database, communication tone rules, the country of the first launch, and the limits of what the agent can do without manual confirmation are needed. !!Automatic cold sales without quality control and provider policies can quickly break the channel!! - this needs to be laid down in the architecture from the start =)
…
Similar examples
> https://business.ingello.com/vorfahr - AI and process automation, close to agency logic and scaling
> https://business.ingello.com/fractal - modular agency architecture and complex processes
> https://business.ingello.com/dent-ingello - medical direction and dental specifics
> https://systems-fl.ingello.com/ua - our profile in system development
-
45 days609 USD
196 45 days609 USD
we have a practically ready foundation for such a multi-agent system - it can be quickly adapted to your funnel, countries, WhatsApp, and local data storage ))
the benchmark for the first working stage is from 180,000 UAH and about 45 days.
this is not the entire cosmos for years ahead, but a normal launch of the core - importing the database, dialogue agent, lead status, local database, control panel, notifications to the owner, basic scenarios for Italy and preparation for Poland or the USA.
we would build this through LangGraph or CrewAI with separate agents for searching, sales, data verification, content, and status control.
important nuance - WhatsApp needs to be done through an official or semi-official provider with careful logic of frequency, templates, and consents, otherwise accounts can quickly face restrictions.
there are questions regarding the technical specifications.
… - what database of dental clinics is already available - format, quantity, country, is there consent for communication
- WhatsApp is needed through Meta WhatsApp API, Twilio, Green-API, or another provider needs to be selected
- the first result is more important as an autonomous sale before the call or as a semi-automatic assistant with manual control by the owner
relevant examples.
- https://business.ingello.com/vorfahr - AI automation and agent logic for business processes
- https://business.ingello.com/fractal - complex architecture of agents and automation
- https://business.ingello.com/lita - medical domain where roles, data, and regulatory nuances are important
main profile on systems - https://systems-fl.ingello.com/ua
I am here, we can discuss the details within the marketplace and quickly gather the first technical plan =)
-
3 days45 USD
738 9 1 3 days45 USD
Hello! I have reviewed your project and am ready to start working. I am confident that I can meet all your expectations.
Current freelance projects in the category AI & Machine Learning
Need an N8N expert for automation and building a system for TikTok farms.
639 USD
My marketing team and I are testing quick ideas for applications. We are looking for a person on a permanent basis to automate processes in N8N, specifically to build a system from scratch and then maintain its functionality. I would appreciate your response! *Unfortunately, I… AI & Machine Learning, System & Network Administration ∙ 2 hours 17 minutes back ∙ 15 proposals |
Development of software with AI
It is necessary to develop software for automatic detection, capturing, and tracking of an object using a video camera and a rotating mechanism.Output data: Video camera with optics. Rotating device on two axes (azimuth/elevation angle). Servomotors controlled via Modbus… AI & Machine Learning, Python ∙ 9 hours 21 minutes back ∙ 10 proposals |
Automation of processes through API and PythonBelow I described the current process and the result I would like to achieve. I also attach files of the real process to better understand how it looks in reality Current process Currently, the entire process is performed manually: uploading/downloading files, transferring… AI & Machine Learning, Python ∙ 14 hours 26 minutes back ∙ 39 proposals |
Finding and Deploying the Best Projects for 4 Mac Mini M4 Machines
229 USD
Finding and Deploying the Best Projects for 4 Mac Mini M4 Machines Hello, I am looking for an experienced technical freelancer who understands Mac Mini M4 hardware, DePIN, blockchain nodes, AI inference, automation systems, Docker, and server monitoring. I currently have 4… AI & Machine Learning, Cryptocurrency & Blockchain ∙ 1 day 1 hour back ∙ 8 proposals |
Development of an AI consultant for a work visa verification website (WordPress)Development of an AI Consultant for a Work Visa Check Website (WordPress) About the Project An English-language website is being developed for the preliminary check of eligibility for a work visa in Israel. An AI consultant in the format of an online chat is needed to assist… AI & Machine Learning, Bot Development ∙ 1 day 9 hours back ∙ 34 proposals |
Truskavets