Skip to content
SOFT-MAN DEVELOPMENT GARAGE

Software,
crafted in the garage.

We build web apps, mobile apps, and the AI inside them: chatbots, LLM integrations, and automation that ships to production. Hands-on engineering, end to end, from first commit to launch and beyond.

See our services Get in touch
2
AI products in house
100k+
workflows automated monthly
-70%
AI cost per conversation

// Why us

Why teams pull into this garage.

AI in production, not in slides.

Our own AI products serve real users every day. We engineer for the parts that decide whether AI survives contact with production: cost per user, failure behaviour, guardrails, and what happens when the model is wrong.

The same discipline goes into client work: AI chatbots for websites and WhatsApp, ChatGPT-style assistants inside your own product, and AI agents that automate real workflows. Everything measured, capped, and monitored.

// AI we work with

AI agents · agentic workflows · MCP · tool use
chatbots · WhatsApp · web widgets · voice AI
RAG · vector search · semantic search · embeddings
multimodal · vision · OCR · speech to text
Claude · GPT · Gemini · open-source LLMs _

Direct access to the engineers.

You work directly with the people writing your code. Decisions get made in one conversation, not relayed through account managers and status meetings.

Fixed scope, fixed price, per phase.

Engagements run in short phases, each with a defined deliverable and budget. You decide after every phase whether to continue, redirect, or stop.

Proven under production load.

The systems we maintain automate more than 100k workflows a month. Everything ships tested, monitored, and documented.

Purpose-built, every time.

No templates, no white-label resells. Each system is designed around your requirements and fully owned by you on delivery.

Long-term support, by default.

Maintenance, security updates, and on-call response are part of the engagement, not a separate negotiation.

// AI Retrofits

AI, bolted on properly.

We fit AI into products the way a good shop fits a turbo: measured, tested, and tuned for the engine it goes into. Our own AI products run in production, so cost per user, failure behaviour, and guardrails are things we engineer, not afterthoughts.

Chatbots for web & WhatsApp

AI customer support chatbots for your website, WhatsApp Business, and Messenger, grounded in your own data. Guardrails, prompt-injection defence, and human handoff built in.

LLM & ChatGPT-style integrations

Claude and GPT wired into your product: summarization, extraction, classification, generation, and assistant features inside your own UI. Structured output with schema validation, behind clean, tested APIs.

RAG & knowledge bases

Retrieval-augmented generation over your documents, wikis, and databases, so answers come from your data, not model guesswork.

AI agents & automation

AI agents and pipelines that take repetitive workflows off your team: support triage, lead qualification, data entry, reporting, follow-ups.

// Common questions

Before you write in.

What kinds of projects do you take on?
Web applications, marketing websites, mobile apps for iOS and Android, APIs, system modernization, and long-term maintenance. We work best on focused projects with a clear owner on your side.
Can you add AI to our existing product?
Yes, that is one of our core services. We integrate LLMs like Claude and GPT into existing systems: chatbots grounded in your data, retrieval-augmented generation over your documents, and automation for repetitive workflows. We treat AI features like any other production code: tested, monitored, with guardrails and fallbacks. And we will tell you honestly when AI is the wrong tool for the problem.
Do you build WhatsApp or website chatbots?
Yes. We build AI chatbots for websites, WhatsApp Business, and Messenger, grounded in your own content and data so they answer accurately about your business. They hand off to a human when the conversation needs one, and we cap and monitor AI spend so running costs stay predictable.
How do you price engagements?
For new builds we usually work in fixed-scope phases of two to four weeks, each with its own budget and deliverable. For maintenance and on-call work we use monthly retainers. We share an honest estimate after a short scoping call, before any contract.
Do you sign NDAs and assign IP?
Yes, both. Standard mutual NDA on the first call if you need it, and full IP assignment on delivery. The code we write for you is yours.
Which timezones do you cover?
We are based in Romania (EET / EEST) and routinely overlap with the United States East Coast, the United Kingdom, and the rest of the European Union. Stand-ups, async-first updates, and dedicated working hours are the default.
Can you take over an existing codebase?
Often yes. We do a paid audit week first to learn the system and write down what we find. After that you decide whether to continue with us, with someone else, or with the current team using our report.
How fast can you start?
For new clients, usually two to four weeks from first contact, depending on current workload. We never overcommit. If we cannot start when you need, we say so on the first call.

// Pull into the bay

Have a project in mind?

Get in touch and tell us about it. We will review your needs and let you know honestly whether we are a good fit for the work.

contact [at] smdevgarage.ro