- Kaptai Rasta Matha, Chittagong 4221
- Mon – Fri: 8:00AM – 6:00PM
- AmzadHossen420360@gmail.com
F2SI LLM Studio is the AI and machine learning division of Friends2a Technology. We design, build and run practical AI for real businesses: agentic apps on leading AI APIs, private models trained on your own data, AI for factory machines, and the consulting, training and support that keep it working.
1Service Line
For software teams, enterprises, e-commerce & SaaS businesses
We build agentic apps and AI-powered software on top of leading large language model APIs such as OpenAI, Anthropic Claude and Google Gemini. Your software can read documents, answer questions from your own knowledge, call your business systems and complete multi-step tasks.
There are no GPU servers to buy or manage. You pay only for what you use, and projects can go live in weeks.
What we build
AI agents for Shopify merchants, delivered through our F2commerce brand.
Smart apps for Wix businesses, delivered through our F2 MIVAQ brand.
Custom AI software built around how your company works.
We add AI features to existing software products for software companies and agencies, from chat and search to full agent workflows.
We connect AI agents to your databases, CRM, ERP and internal tools through MCP servers and function calling, with clear permissions on what each agent can do.
Every system ships with guardrails, accuracy testing and monitoring, so you can see how it performs and what it costs.
Technologies
2Service Line
For security-sensitive companies, finance, legal, healthcare & e-commerce
We fine-tune open-source models such as Llama, Qwen, Gemma and Mistral for your specific tasks, and build small custom models from your own data. The models run on your own servers or private cloud, so sensitive data never leaves your control.
A small model trained for one job can be fast, accurate and far cheaper to run at high volume than a large general model.
What we build
Chatbots and knowledge assistants running on open-source models inside your own servers, answering from your documents.
Small models trained with LoRA for one job: extracting data, classifying messages or writing summaries in your format.
OCR and extraction models that read invoices, forms, contracts and scanned documents and turn them into structured data.
Models that detect objects, count items and recognise defects in images and video, built with YOLO and PyTorch.
Forecasting sales and demand, scoring risk and detecting fraud from your historical business data.
Product recommendations and smart search for e-commerce stores and internal knowledge.
Quantization and tuning that make models smaller and faster, so they run on cheaper hardware and edge devices.
Technologies
3Service Line
For factories, manufacturing plants, warehouses & production lines
We bring AI to factory machines and production processes. Cameras and sensors on the floor feed AI models running on local edge devices, which spot defects, warn before machines fail and show managers what is happening in real time.
For work that controls machines directly, such as PLC and robotics integration, we work alongside industrial automation partners and control engineers.
What we build
Cameras and vision models that check products on the line for defects in stitching, colour, labels, caps or packaging.
Sensor data on temperature, vibration and power analysed to warn your team before a machine breaks down.
Live dashboards for output, uptime, downtime and efficiency, with daily summaries for managers.
CCTV-based detection of helmets and safety vests, and alerts when people enter restricted areas.
Automatic counting of products on conveyors and stock in warehouses, and tracking of trucks and loads.
Analysis of electricity, gas and boiler usage to find waste and lower energy costs.
Connecting AI results to PLC, SCADA and robots, for example to reject a faulty product automatically, delivered with automation partners.
Technologies
4Service Line
For business owners, management teams, IT departments & universities
Not every problem needs AI, and not every AI project needs a custom model. We help you find where AI truly saves time and money, choose the right approach and plan the cost before you build.
We also train your teams to use AI tools in daily work, and train developers to build AI systems themselves.
What we offer
We review your processes and data and deliver a report on where AI can save time and cost, and where it cannot.
Which projects to do first, API or private models, expected costs and a realistic timeline.
Training for staff and managers to work faster with AI tools for reports, email, analysis and research.
Hands-on courses in AI engineering (APIs, RAG, MCP, agents) and machine learning (fine-tuning, evaluation).
One to three day practical workshops for universities, incubators and industry groups.
Clear rules on what information staff may share with AI tools, to protect data and privacy.
Methods & Tools
5Service Line
For every company running AI in production
Launching an AI system is only the start. We keep it reliable month after month: watching performance, updating models, controlling costs, keeping data secure and helping your users.
Support plans cover the systems we build and can also cover AI systems you already run.
What we provide
Setting up AI systems on your servers, cloud accounts or edge devices with Docker and proper configuration.
Tracking uptime, speed and answer quality, with alerts when something changes.
Adding new documents and data, retraining custom models and moving to better models as they appear.
Watching API and server spend, routing simple work to cheaper models and removing waste.
Access control, data protection, logging, backups and recovery plans.
Help for your team when issues come up, small changes and training for new users.
Technologies
How We Work
Every project follows the same clear path, so you always know what comes next and what it costs.
We study your workflow, data and goals, and agree on one clear problem to solve.
We choose the right approach, API, private model or both, with scope, cost and timeline.
We build a working pilot, test it on your real data and measure the results.
We deploy into your systems, connect your tools and train your team.
We monitor, update and improve the system as your business grows.
Deployment Options
We match the setup to your data rules, budget and speed needs.
Leading models from OpenAI, Anthropic and Google. No servers to manage, pay per use.
Models deployed in your own AWS, Azure or Google Cloud account, billed to you.
Open-source models on your own servers. Data never leaves your building.
Small models on devices such as NVIDIA Jetson for real-time, offline results.
Tell us about your business and the problem you want to solve. We will suggest a practical starting point and a clear plan.