Home / AI Agents
Any AI agent, built to your brief

Agents thatnever sleep.

If you can describe the job, I can build the agent. That goes from a single chatbot to a coordinated team of agents, on any channel, in Bangla or English.

7Production AI systems
~15KFree API requests/day
3LLM providers
24/7Always on
What I build

Not limited to one
platform or tool.

n8n is one of the tools I use, not a limit. I pick the right stack for each job.

💬

Customer-support & sales

Answer questions, qualify leads, book calls and hand over to a human on WhatsApp, Messenger, Telegram, web chat and email.

🎙️

Voice & call agents

Voice notes, phone-style booking calls, speech-to-text summaries and spoken replies.

📚

Knowledge & RAG

Grounded in your documents, SOPs and databases, with sources and a safe "I don't know" fallback.

🔭

Research & analyst

Scheduled agents that browse, compare and report on markets, competitors, prices and leads.

🧾

Document & data

Read invoices, forms, PDFs and emails, then extract, validate and push clean data into your systems.

🛠️

Coding & DevOps

Code review, bug triage, uptime and log monitoring. Agents that diagnose problems and open the ticket.

📣

Marketing, ads & SEO

Content pipelines, social posting, comment moderation, ad audits and SEO reports, every day.

🗂️

Back-office & ops

Orders, invoices, inventory, HR screening, approvals and reminders, automated.

🔌

Tool-using agents

Call APIs, read and write inside your apps, fill forms and trigger other systems (REST, webhooks, MCP).

🕸️

Multi-agent teams

An orchestrator with specialist agents (researcher, writer, checker, publisher) that merge their results.

🏥

Education, health & NGO

Multilingual information, appointment and follow-up agents with strict data control.

✨

Something not on this list?

Describe the job in plain words. If it follows a process, we'll scope the agent, the stack and the cost together.

GPT-4 / 4oGeminiClaudeGroqOpenRoutern8n · self-hostedPython · FastAPIPinecone RAGMeta Graph APIWhatsAppTelegramVoiceDocker
Deployed production systems

Real automations,
real results.

Pick a system to see how it works, step by step.

Live workflows

Step by step,
end to end.

Five production workflows exactly as they run today. Scroll and each diagram powers on, node by node.

Workflow blueprints

Buildable today.

Seven advanced architectures, multi-step, multi-agent and human-in-the-loop, on the same self-hosted stack.

Multi-agent · 7 stages

✍️ AI Content Factory

Researches, writes, fact-checks, designs and publishes across your channels, with your approval before anything goes live.

Topic radarResearchWriterFact-check & SEOImageApprovalPublish
Sales · 7 stages

🎯 Lead-to-Close Sales Agent

Every lead is enriched, scored, answered within seconds and followed up until it books a call.

Lead formEnrichScoreInstant replyBookingCRMFollow-ups
Support · 7 stages

🛟 Omnichannel Support with RAG

Answers only from your approved knowledge, and hands over to a human with a summary when unsure.

Message inIntentRAG searchAnswer + sourcesConfidenceHandoffTicket + CSAT
Voice · 7 stages

📞 Voice Appointment Agent

Takes the call, understands the need, books the appointment and confirms it by message, day or night.

Inbound callSpeech-to-textReasoningCalendarBookConfirmLog
Back-office · 7 stages

🧾 Document & Invoice Agent

Documents arrive by email or chat. Clean, validated data comes out, and only real exceptions reach a person.

InboxClassifyExtractValidateExceptionsPush to systemReport
Research · 6 stages

🔭 Market & Competitor Research

Scout agents gather data every morning, an analyst compares it, and you get a one-page decision brief.

ScoutsCollectCleanAnalyseBriefDeliver
DevOps · 7 stages

🛠️ Self-Healing DevOps & Monitoring Agent

Watches your servers and sites, diagnoses what broke and proposes the fix. After your approval, it runs the safe action and writes the incident report.

MonitorDetectDiagnosePropose fixApproveRun actionIncident report
FAQ

Questions, answered.

Can you build any kind of AI agent?

Yes: any process that can be described. That includes support, sales, voice, research, document processing, back-office, marketing, coding and DevOps, and multi-agent teams.

I don't know exactly which agent I need. Can you help?

Yes. Tell me the task your team repeats most. In a free discovery call we turn it into an agent design, the right stack, a timeline and a clear price.

Do you only work with n8n?

No. n8n is my main tool for fast, self-hosted automation, but I also build custom agents in code (Python / FastAPI), connect any API or webhook, and combine several LLMs when needed.

What does "self-hosted" mean for me?

Your agents run on your own server (a VPS with Docker + n8n) instead of a monthly SaaS platform. You keep control of your data, pay only for hosting and API usage, and avoid per-seat fees.

Which channels can the agents work on?

Facebook Messenger, WhatsApp, Telegram, web chat, email and voice-note workflows, with a unified inbox and human handoff.

Can the agent reply in Bangla?

Yes. My Messenger agent replies in fluent Bangla, and agents can also work in English and other languages.

Which AI models do you use?

Google Gemini, GPT-4 / 4o, Claude, Groq and OpenRouter, chosen per use case for quality, speed and cost. Company knowledge is added through Pinecone RAG.

How long does it take to build an agent?

A focused single-channel agent takes about one to three weeks. Multi-channel or multi-agent systems take longer.

Do you offer support after launch?

Yes: monitoring, optimisation and new workflows as an ongoing retainer, plus hosting and server management.