Highload system

Budget: 2500 USD Deadline: 21 days

Hello. I will design and deploy a high-load architecture for the backend in Python (FastAPI, Celery/Redis, PostgreSQL/VectorDB) with integration of multimodal models (CLIP / Gemini Flash / Qwen-VL) and Telegram/WhatsApp API.

Vector search & LLM: For searching by photo and text simultaneously, I will implement the generation of unified embeddings stored in Qdrant / Milvus. This will ensure the delivery of a clothing retail catalog with sub-second response time.

Budget: 2500 USD Deadline: 45 days

I will take the first production-MVP of multimodal search: uploading photos and text, combined search, LLM ranking, API, and integration with a messenger through a personal assistant.

I will divide the architecture into a search service, vector index, processing queue, and agent layer — to further scale the load without reworking the core.

Deadline: 45 days. Cost of the first stage: $2,500.

Is the clothing catalog and labeled data already available through API or export?

Budget: 2500 USD Deadline: 60 days

Good day.
I am implementing a high-load backend on NestJS with integration of LLM and multimodal search by photo and text, as well as connecting a personal AI assistant to messengers. To work with multiple AI models, I suggest using OpenRouter.ai, which will allow easy switching between models and scaling the system without being tied to a single provider.
I am ready to discuss the details and start working. I would be happy to collaborate.

Budget: 2500 USD Deadline: 30 days

Hello.

I am a NodeJS developer. I am ready to take on the task. Write to me, and we will discuss.

Budget: 2400 USD Deadline: 14 days

Good day, over 5 years in development, complex high-load services with AI on SaaS - that's my specialty.

I created my project Throngwatch independently: it's a SaaS analytics tool with AI reports for Discord, full stack from architecture to launch.

We can start with an MVP: multimodal clothing search (photo + text), LLM integration, personal assistant agent in a messenger, well-thought-out backend for high load. The entire stack - from scratch, Node.js/React, deployment for traffic. The basic version is estimated to take about 2-3 weeks; the exact cost after a few clarifications or in personal messages.

Which messengers are a priority for integration? Is a separate admin panel needed for management and analytics?

I look forward to your message in personal!

Budget: 2500 USD Deadline: 14 days

The budget of 2500 USD I would allocate for the first engineering phase - architecture, technical model, prototype of multimodal search, and evaluation of the production version. A complete high-load system with LLM, search by photo and text, an agent in messengers, queues, vector database, monitoring, and proper security will be a separate phase and, most likely, significantly higher in budget.

In terms of implementation, I would focus not on a beautiful demo, but on the quality of search and cost per query. First, we check the pipeline - product photo, text query, embeddings, ranking, filters, agent responses, limits, and scenarios in messengers. Then we package this into a stable architecture with queues, caching, logging, admin panel, and quality metrics. Otherwise, you can quickly end up with an expensive toy that searches by inspiration, and the bill for the model also has character =/

Questions for accurate assessment
> what clothing catalog is planned at the start - number of products, photos, attributes, update frequency
> which messengers are needed - Telegram, Viber, WhatsApp, Instagram, Facebook Messenger, or another set

Similar cases Ingello
> https://business.ingello.com/vorfahr - AI and process automation with agent logic
> https://business.ingello.com/fractal - complex AI architecture and system automation
> https://business.ingello.com/carveli - e-commerce logic, catalog, and product system

Main profile for FLH
> https://systems-fl.ingello.com Expand

Budget: 2500 USD Deadline: 7 days

Hello, I recently created a similar service with search and a Telegram agent: advanced search with filtering, Telegram API integration, role-based access, notifications, and an admin panel. A multimodal scenario, where the request comes in with both text and an image, is layered on top of this search with an LLM layer.

If the assortment is large and growing, it's worth immediately implementing a vector index for both photos and text, as this keeps the response time for searches stable under load.

How many items are in the catalog at the start and what is the expected peak query flow? And where do the product images come from, is it an in-house database or parsing from external sources?

I suggest we get in touch; I will also sketch out a solution scheme for multimodal search and agents in messengers, with a breakdown for the LLM and high-load parts.

Budget: 2500 USD Deadline: 30 days

Good day!

I can develop a high-load service that will find clothing in the catalog based on photos and text, working as a personal agent directly in Telegram or any other messenger.

The core itself — photo search — I have already assembled: a Telegram bot with CLIP visual search for the product catalog (photo input → similar items from the catalog).

What else I cover:
- High-load — I architected a production system for 2.5M orders / 14K+ active users per month with real-time (taximeter and active order);

I separately deployed ML models on GPU infrastructure (VastAI, Kaggle, multi-GPU) as services.

- Multimodality/CV — I orchestrated 6 foundation models (CLIP, SAM, Florence-2, Grounding DINO, Flux) on fashion images, not just basic CLIP lookup.

- Agent + LLM — multi-agent systems from scratch, orchestration of Claude/Mistral/Qwen/GPT-OSS; separately an LLM pipeline on the clothing marketplace.

A couple of questions to assess the specifics:
- Scale — size of the catalog and requests/sec at peak?
- Are photos and text ranked together (hybrid), or photo → candidates → text?
- Which messengers, and is the catalog yours or external?

The system is large, so I would consider it step by step; the exact range — after discussing all the details.

I work through Business Safe. Expand

Budget: 2500 USD Deadline: 30 days

I studied the task: multimodal search by photo and text simultaneously, integration into messengers, LLM agent. This is an interesting and non-trivial system where it is important to properly organize the pipeline from the incoming user request to the agent's response.

I have about 3.5 years of commercial experience in full-stack development, including work on B2B e-commerce and products with high architectural requirements. Stack: TypeScript, NestJS, PostgreSQL, Vue/Nuxt, React/Next.js. I understand how to build modular systems with DI, where components (vector search, LLM orchestration, messenger adapters) do not intertwine and are easily scalable.

Regarding the high-load part: it is critical to properly establish task queues and caching between the embedding model and the main LLM to avoid bottlenecks under load. For integration into messengers: it is better to design through a single adapter layer, so adding a new messenger does not break the agent's logic.

I am ready to discuss the details of the architecture and scope of work.

Budget: 3000 USD Deadline: 30 days

Hello! I will be developing a high-load system with LLM for multimodal clothing search: photo + text, with integration into messengers through a personal assistant bot. I have experience with high-load solutions involving AI integration and working with messengers.

Which LLM platform do you plan to use for recognition and semantic search based on clothing photos?

Budget: 2500 USD Deadline: 27 days

Hello, I am ready to develop a high-load system for multimodal clothing search by photo and text query with integration of LLM and personal agent in messengers. I have relevant experience in Python backend, AI/LLM integrations, APIs, queues, databases, and building services considering load. I envision the implementation through Python/FastAPI for the backend, a separate layer for processing images and text queries, a task queue, caching, and an API for connecting messengers. First, I will finalize the architecture, search scenarios, and load requirements, then I will build the MVP with further optimization for speed, stability, and cost of requests to the models. For FastAPI/Python backend, I developed the backend for the project: https://zem.center/ git: github.com/onyx144

Budget: 2500 USD Deadline: 3 days

Hello! I have reviewed the project and am ready to start working. I am confident you will be satisfied with the result.

Budget: 2500 USD Deadline: 7 days

Good day, Nick!

I have experience in developing high-load systems and integrating with AI. This allows for the creation of effective solutions for multimodal search that meet the needs of the modern market.

I understand that you are working on creating an online service for clothing search that combines text queries and images. This requires a deep analysis of the target audience and technologies to ensure a high level of user convenience.

To form a detailed strategy, I propose the following steps:
1. Familiarize yourself with the technical requirements and architecture of your system.
2. Identify the key features that need to be implemented for integration with messengers.
3. Conduct a competitor and market analysis to understand what solutions already exist.

The preliminary cost of my services starts from $400 per month. Depending on the complexity of the tasks, the implementation timelines may vary from 2 to 8 months.

In my profile, you can review examples of completed projects: Freelancehunt

I suggest discussing your project in private messages to explore the goals and plans for development in more detail. Expand

Budget: 2500 USD Deadline: 30 days

Hello! What is described is actually three big tasks at once: high-load infrastructure, multimodal search (photo + text simultaneously), and an agent-assistant in a messenger. For $2500, it cannot be fully implemented — and to be honest, discussing "high-load" is premature until there is real load; it's not the right place to start.

What I propose in fact: an MVP on the existing multimodal model (embeddings for photos and text, vector search for your catalog) plus a Telegram bot on top with LLM for understanding the request — without training your own model and without high-load architecture, which is not needed yet. I have already written a similar setup (hybrid search, embeddings, vector database) by hand in my project.

Please tell me about your catalog — how many products, whether there are already photos and descriptions in a structured format — and I will provide a specific price and timeline for the first stage.

Budget: 2500 USD Deadline: 10 days

Hello!

I looked at the project description. I already have an understanding of how it can be implemented.

I would build the system using multimodal embeddings to simultaneously consider photos and text, and the search would work quickly even with a large catalog. I would use LLM as an intelligent assistant in messengers, and the search itself would be moved to a separate service so that it could be scaled independently.

Before development, I will propose an architecture to ensure proper performance from the start and avoid problems as the load increases.

If there are already preferences regarding the stack or any existing developments, I can adapt. If not, I will help choose the optimal solution.

I would be happy to discuss the project details. Expand

Budget: 2500 USD Deadline: 14 days

We already have a nearly ready similar solution for AI agents and search engines, which can be quickly adapted for clothing and messengers. I suggest discussing the details here; I'm available.

Regarding timelines, a working technical prototype can be created in 10-14 days. I would include multimodal search by photo and text, vector search, a basic product showcase, a personal agent, and integration with the first messenger.

In terms of architecture, I would divide the system into product ingestion, image processing, embeddings, search, the LLM layer of the agent, task queue, and an admin part for quality control of the output. This way, it will be easier to scale the load and not rewrite everything from scratch.

We need to clarify two points:
- What product catalog and how many items are expected at the start?
- Which messengers are needed first - Telegram, Viber, Instagram, WhatsApp, or another channel?

Similar cases:
- https://business.ingello.com/vorfahr - AI and automation of business processes
- https://business.ingello.com/fractal - agent architecture and complex scenarios
- https://business.ingello.com/prime-eva - e-commerce and systematic work with product data

Our profile for such systems - https://systems-fl.ingello.com Expand

Budget: 2500 USD Deadline: 3 days

Multimodal search is a combination of a CLIP-type model for embedding photos and text and a Qdrant vector database, which quickly finds similar products under load.

I will index the clothing catalog, process photo and text queries simultaneously, implement queues and caching for high load, and create an agent on LLM for dialogue in the messenger to clarify size and color. The search prototype will be ready in 3 days.

Please let me know which messenger to use first and if there is already a product catalog with photos.

Budget: 2500 USD Deadline: 14 days

Hello! In multimodal search, it is critically important to properly configure the caching of vector embeddings so that the LLM does not overload the GPU with every clothing request. I have created similar distributed systems.

What database do you plan to use for vector search (e.g., Pinecone or Milvus)?

Let's get in touch, and I will propose an architectural plan for load optimization.