< agent lab />

Inside the control room.

This is how I think about agentic AI. Every system I ship follows a similar shape: an orchestrator brain, a typed set of tools, a memory layer, and a channel for the user to talk to it.

< generic agent architecture />

How an agent thinks

USER INPUT

text · image · voice

→

orchestrator

AGENT BRAIN

OpenAI Agents SDK · Gemini

→

OUTPUT

WhatsApp · Streamlit · API

tool

Web Search

Live web retrieval via SERP APIs

Tavily · SerpAPI

tool

RAG Knowledge Base

Cited answers from documents

LangChain · Pinecone

tool

Code Execution

Sandboxed Python for analysis

Python REPL

tool

External API Call

Connect to any HTTP service

n8n · Fetch

tool

Human Escalation

Route to a human when uncertain

WhatsApp · Slack

< interactive demo />

Try it yourself

Pick a scenario and watch a representative execution trace step by step. These demos mirror the architecture patterns used in the shipped projects below.

Input

"What is the attention mechanism in transformers?"

○

Embedding Query

○

Vector Search

○

Reranking Results

○

Generating Answer

< recorded traces />

Agent runs you can reason about.

Live demos are risky for public agent systems, so this page uses safe recorded traces and representative runs. Each trace highlights the decision points a recruiter or client should inspect.

safe demo mode

recorded

Selective retrieval trace

The agent decides whether retrieval is needed before calling document tools.

intent checktool decisionretrievalcited response

recorded

Crop diagnosis trace

A vision result is converted into treatment guidance and farmer-friendly next steps.

image intakevision diagnosistreatment planlocalized reply

design trace

Personal ops trace

Events become prioritized tasks with human approval before sensitive actions.

watchertask structuringapproval gatebriefing

< case studies />

Agents I've shipped

View all →

AI / Agent

Personal AI Employee

Autonomous AI Agent with Claude Code & MCP

RESEARCH

Multi-Channel Autonomous AI Workflow

Autonomous AI agent system that manages real personal and business workflows. Monitors multiple communication channels (Gmail, WhatsApp, LinkedIn) using watcher scripts and convert...

Proof

Multi-channel event monitoring; Autonomous workflow management; Human-in-the-loop approvals; Live observability das...

Multi-channel IntakeEvent StructuringTask PrioritizationVault Routing+3

Claude CodeMCPObsidianGmail APIWhatsApp API+3

Case Study Post

AI / Agent

Zorox

WhatsApp Agent w/ Intent Routing

RESEARCH

Intent Routing Pipeline

Production WhatsApp agent with multi-intent routing, context memory, and human escalation paths.

Intent ClassifierContext MemoryTool SelectionEscalation

n8nGeminiRedisPostgres

Case Study GitHub

AI / Agent

OpenClaw

Personal AI Employee Platform

LEARNING

AI Employee Registry

A platform of specialized AI employees you can hire on demand — research, ops, content, and code agents collaborating via ACP.

Hiring UIAgent RegistryTask OrchestratorTool Bus

OpenAI Agents SDKACPMCPNext.js

Case Study GitHub

AI / Agent

n8n WhatsApp AI Agent

Multi-Modal Automation with Google Gemini

LIVE

Multi-Modal Workflow

Sophisticated n8n workflow creating an intelligent WhatsApp bot with multi-modal capabilities. Processes text, voice, and images using Google Gemini AI. Features smart routing for...

Proof

Handles text, image, and voice inputs in one documented workflow

Message TriggerType DetectionMedia DownloadAI Processing+2

n8nWhatsApp APIGoogle GeminiWorkflow AutomationMulti-Modal AI

Case Study GitHub

AI / Agent

Multi-Agent Chainlit Assistant

OpenAI Agents SDK with dynamic routing

LIVE

Multi-Agent Orchestration

Intelligent assistant framework using OpenAI Agents SDK with Chainlit UI. Features a main routing agent that intelligently dispatches queries to specialized sub-agents: Shaitani Ca...

Proof

Autonomous agent routing with specialized tool handling

Query InputMain RouterAgent SelectionTool Execution+1

OpenAI Agents SDKChainlitPython 3.11uvdotenv

Case Study GitHub

AI / Agent

AgriSahayak

AI-Powered Crop Disease Detection Agent

LIVE

Disease Detection Vision

WhatsApp-based farming assistant using Gemini Vision AI for real-time crop disease detection, treatment plans, and weather alerts. Bilingual Urdu/English.

Proof

60% projected crop loss reduction

Vision InputDisease ClassifierTreatment PlannerAlert System

Gemini Visionn8nWhatsApp APIPython

Case Study GitHub

View all agents →

< tech ecosystem />

How the pieces connect

Channels

WhatsApp Business API
Streamlit
REST API
CLI

Orchestration

n8n
OpenAI Agents SDK
Custom Router
MCP / ACP

Models

Google Gemini (text + vision)
OpenAI GPT-4 / 5
HuggingFace OSS
Whisper (ASR)

Knowledge

LangChain
Pinecone
ChromaDB
Postgres pgvector

Tools

Web Search
Code Execution
File I/O
External APIs

Memory

Redis (short-term)
Postgres (long-term)
Conversation summarization