Skip to content
View asadullah48's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report asadullah48

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
asadullah48/README.md

Asadullah Shafique — Agentic AI Developer · CMT Industry Founder · Digital Marketing Strategist

Typing SVG




 Who I Am

Agentic AI Developer at Panaversity — I build autonomous AI systems that take action, not just respond.

Completed 6 progressive hackathons evolving a simple file watcher into a Kubernetes-orchestrated platform with Constitutional AI safety, Kafka event streaming, Dapr service mesh, and a Discord bot.



 At a Glance

role        : Agentic AI Developer @ Panaversity
hackathons  : 6/6 completed (Bronze → Platinum)
k8s_cluster : 14 services · 6GB · 44% memory
tests       : 180+ passing
methodology : Specification-First Development
framework   : Harness × Loop × Graph engineering
focus       : Multi-Agent Systems (MCP + A2A)

 Currently Building

  01 Multi-agent systems with MCP & A2A protocols
  02 Production RAG with vector databases
  03 Platform engineering for AI workloads
  04 Open-source AI safety tooling
  05 Technical content on YouTube & Medium



🏭  The Agent Factory Vision

In the AI era, the most valuable companies won't sell software — they'll manufacture AI employees, powered by agents, specs, skills, MCP, autonomy and cloud-native technologies.

The shift from developer-as-typist to developer-as-orchestrator is here. I'm building systems where natural-language specs drive autonomous agents that don't just respond — they act, coordinate, and deliver.

Concept Description
Digital FTEs AI agents that function as full-time digital employees, handling end-to-end workflows
Agent Skills Reusable, composable capabilities (39 skills across 8 categories on Claude.ai)
Spec-Driven Automation From manual coding to specification-first development — the spec IS the product
Agent Protocols MCP (Model Context Protocol) & A2A (Agent-to-Agent) for standardized agent communication

📊 Reference: Agent Factory: Building Digital FTEs — Presentation



🧭  How I Build Agents — Harness × Loop × Graph

A model is not a product. The harness, the loop, and the graph around it are.


Pillar Discipline The Principle
01 Harness Engineering The Environment Tools, memory, permissions, observability. An agent that forgets what it did five steps ago is a harness problem, not a model failure.
02 Loop Engineering The Feedback Evidence over confidence. A coding agent stops when the tests pass — not when it feels done.
03 Graph Engineering The Flow Branching, approvals, retries, parallel tasks. One agent is a demo; a graph with approval gates is a product.


 Flagship Project


A production-grade, Kubernetes-orchestrated AI platform built across 6 progressive hackathons — featuring Constitutional AI safety, event-driven microservices, and multi-interface access.


%%{init: {'theme': 'dark', 'themeVariables': {'primaryColor': '#1f6feb', 'primaryTextColor': '#e6edf3', 'lineColor': '#30363d', 'secondaryColor': '#161b22', 'tertiaryColor': '#0d1117'}}}%%
graph TD
    subgraph "🎮  User Interfaces"
        A["🎮 Discord Bot<br/><sub>TodoMaster AI</sub>"]
        C["🌐 Next.js Frontend<br/><sub>React Dashboard</sub>"]
    end

    subgraph "⚙️ &nbsp;Core Platform"
        B["⚡ FastAPI Backend<br/><sub>+ Dapr Sidecar</sub>"]
        I["🛡️ Constitutional AI<br/><sub>Safety Middleware</sub>"]
    end

    subgraph "💾 &nbsp;Data Layer"
        D["🗄️ PostgreSQL 15<br/><sub>StatefulSet</sub>"]
        H["📦 Redis 7<br/><sub>State Store</sub>"]
    end

    subgraph "📡 &nbsp;Event Streaming"
        E["📡 Apache Kafka<br/><sub>Strimzi KRaft</sub>"]
        F["🔔 Notification<br/><sub>Service</sub>"]
    end

    subgraph "📊 &nbsp;Observability"
        G["📊 Prometheus<br/><sub>Metrics</sub>"]
    end

    A -->|REST API| B
    C -->|REST API| B
    I -->|Block · Flag · Allow| B
    B -->|SQL| D
    B -->|State| H
    B -->|Dapr Pub/Sub| E
    E --> F
    G -->|Scrape| B
    G -->|Scrape| F

    style A fill:#5865F2,color:#fff,stroke:#5865F2,stroke-width:2px
    style C fill:#000000,color:#fff,stroke:#30363d,stroke-width:2px
    style B fill:#009688,color:#fff,stroke:#009688,stroke-width:2px
    style I fill:#EF4444,color:#fff,stroke:#EF4444,stroke-width:2px
    style D fill:#336791,color:#fff,stroke:#336791,stroke-width:2px
    style H fill:#DC382D,color:#fff,stroke:#DC382D,stroke-width:2px
    style E fill:#231f20,color:#fff,stroke:#30363d,stroke-width:2px
    style F fill:#4ECDC4,color:#fff,stroke:#4ECDC4,stroke-width:2px
    style G fill:#e6522c,color:#fff,stroke:#e6522c,stroke-width:2px
Loading
Key Differentiators
Feature Implementation
Constitutional AI Blocks homework-solving queries with Socratic responses, flags edge cases for human review
Zero-Code Infra Swap Switched pub/sub from Redis → Kafka by changing 1 YAML file (Dapr abstraction)
14 Services in 6GB Full production stack at 44% memory utilization on Minikube
Event-Driven Audit Every interaction published to Kafka with 24h retention
Multi-Interface Same backend serves Next.js frontend + Discord bot (TodoMaster AI)
Hackathon Progression — Bronze → Platinum
# Project Tier What I Built Tests
0 Personal AI CTO Bronze File watcher, auto-categorization, HITL approvals 7/7
1 Course Companion Silver FastAPI backend, Constitutional AI filter, conversation tracking
2 AI-Powered Todo Silver Spec-driven development, AI spec generation, CRUD with constitution
3 Advanced Todo Gold Event-driven architecture, Kafka, Dapr, team collaboration 149/149
4 Cloud-Native Platinum Full Kubernetes cluster (14 manifests), CI/CD, Prometheus
4.5 Discord Bot Extended TodoMaster AI with 6 slash commands, K8s deployment 31/31


🔧  Engineering Expertise

Production-ready command across backend engineering, infrastructure, cloud, and AI — sharpened through 6 hackathons and live deployments.

Domain Expertise
Python Core Syntax & Data Structures · OOP & Design Patterns · Async/Await & Concurrency · Type Hints & Decorators · Testing & Debugging · FastAPI Integration
Docker & Containerization Docker Architecture · Images & Container Lifecycle · Dockerizing Node.js Apps · Port Mapping & Networking · Docker Compose · Docker Networking · Persistent Volumes
Redis & Caching Redis Data Structures · API Response Caching · Rate Limiting · Message Queue with Redis Pub/Sub
System Design & Scalability System Design Principles · Horizontal & Vertical Scaling · Nginx (Reverse Proxy & Load Balancer) · Microservices Architecture · DB Replication & Sharding
CI/CD & Cloud Infrastructure CI/CD Pipeline Design · AWS Deployment & Services · Infrastructure as Code (IaC)
AI Integration in Backend LLM APIs & Prompt Engineering · LangChain Framework · Retrieval-Augmented Generation (RAG) · Vector Databases (Pinecone, Weaviate, Milvus)


🏭  Founder — Texcot Embroidery Sourcing House

Built and deployed a production SaaS (CMT Stitching System) for the garment industry — because I run one. Order management, billing, inventory, dispatch, and financial tracking. The only CMT software built by a CMT owner.





🚀 Latest Ship — Textile ERP Platform

Live Repo Backend DB Tests

A production-deployed, multi-tenant SaaS for Fabric Mill inventory management — built spec-first across one session from zero to live in under 4 hours.

Architecture : Multi-Tenant SaaS with PostgreSQL Row Level Security
Backend      : FastAPI + asyncpg + Alembic (19 API routes)
Frontend     : Next.js 15 + shadcn v4 + TypeScript strict
Auth         : JWT with tenant-scoped sessions
Infra        : Koyeb (backend) · Vercel (frontend) · Neon (DB)
Safety       : RLS enforced at DB level — tenants cannot see each other's data
Tests        : Tenancy isolation suite (two-tenant cross-contamination checks)
Deployed     : ✅ Live in production
What it does
Feature Detail
Multi-Tenancy Row Level Security on every table — one DB, zero data leaks
Fabric Lot Management Create, track, and manage fabric lots with full CRUD
Roll Tracking Nested fabric rolls per lot — length, weight, status, location
Dashboard Analytics Live stat cards: total lots, meters available vs. reserved
Tenant Registration Self-serve onboarding — company name → isolated workspace in seconds
Docker Compose Full local stack (Postgres + FastAPI + Next.js) with one command

🚀  Live Projects

🤖 Agentic OS Portfolio — A Portfolio That Is Itself an Agent System

Portfolio AI Chat Stack

My personal portfolio built as an agentic system — a multi-mode AI chat agent (Portfolio Guide, Backend Specialist, Frontend Architect, Agent Builder) with streaming SSE, backed by FastAPI on Render, plus the Harness × Loop × Graph agent-engineering showcase.

Site         : Agentic OS Portfolio — asadullahshafique-devunity.vercel.app
Stack        : Next.js 15 · TypeScript · Tailwind CSS · FastAPI · Render
Agent        : 4-mode chat agent · streaming SSE · Discord-webhook contact pipeline
AI-Ready     : llms.txt · JSON-LD Person schema · dynamic OG image
Languages    : English + Arabic (full RTL)
Deployed     : ✅ Live on Vercel (frontend) + Render (backend)

🧵 CMT Stitching System — Garment Production Management

Live App Stack Industry

A garment production tracking system for CMT (Cut, Make & Trim) operations — managing stitching orders, packing, and production workflows for the textile manufacturing industry.

Use Case  : CMT Stitching & Packing Management for garment factories
Stack     : Next.js · TypeScript
Live      : ✅ Deployed on Vercel
Domain    : Textile & Garment Manufacturing (Pakistan · UAE)

🌐 Dubai E-Marketing & Business Services

Services Industry Region

Beyond software, I provide digital marketing and e-marketing solutions for Dubai-based businesses across construction, trading, and import/export sectors — and for Pakistani SMEs entering the digital economy.

Service Description
E-Marketing Strategy Digital presence, SEO, and online lead generation for UAE markets
Construction Sector Marketing for contractors, fit-out companies, and building material suppliers
Trading & Import/Export Online brand building for commodity traders and international sourcing agents
Pakistani SMEs AI tools, digital marketing, and SaaS solutions for Pakistani factories and businesses entering the digital economy
AI-Powered Automation Automated marketing workflows using AI agents — from content to outreach
Product Sourcing Textile and garment sourcing with production management systems

📧 For business inquiries: texcotembroiderysourcinghouse@gmail.com 🔗 Connect: LinkedIn · Linktree



 Other Projects

  Project Stack Description
01 Physical AI Textbook Platform Next.js · FastAPI · RAG · Gemini Interactive textbook with semantic search and context-aware RAG chatbot
02 LearnFlow AI Platform Microservices · FastAPI · K8s · Docker 5 specialized AI agents for personalized programming education
03 Course Companion FTE FastAPI · ChatGPT API · Zero-Backend Constitutional AI rules for LLM-based course management
04 Claude.ai Skills Marketplace 39 Skills · 8 Categories Reusable agent skills — doc processing, automation, dev tools


 Tech Stack

LANGUAGES & FRAMEWORKS

Languages

CLOUD-NATIVE & INFRASTRUCTURE

Infrastructure

AI · TOOLS · PLATFORMS

AI & Tools

Full Stack Breakdown
const techStack = {
    languages      : ["Python", "TypeScript", "JavaScript"],
    frontend       : ["Next.js 15", "React", "Tailwind CSS"],
    backend        : ["FastAPI", "Node.js", "Uvicorn"],
    ai_ml          : ["Constitutional AI", "RAG Systems", "LangChain", "LangGraph", "MCP"],
    databases      : ["PostgreSQL 15", "Redis 7", "Pinecone", "Chroma", "Neon (serverless Postgres)"],
    deployment     : ["Vercel (frontend)", "Koyeb (backend)", "Neon (database)"],
    infrastructure : ["Kubernetes", "Docker", "Dapr", "Helm"],
    streaming      : ["Apache Kafka (Strimzi KRaft)"],
    monitoring     : ["Prometheus", "Grafana", "OpenTelemetry"],
    cicd           : ["GitHub Actions — test → build → validate → security"],
    bots           : ["discord.py (slash commands)"],
    apis           : ["OpenAI", "Claude (Anthropic)", "Google Gemini"],
    protocols      : ["MCP", "A2A", "REST", "Dapr Pub/Sub"],
    architecture   : ["Microservices", "Event-Driven", "API-First"],
    methodology    : "Specification-First Development"
};


🌐  Domain Expertise

Textile & CMT Manufacturing · Dubai Real Estate Marketing · Digital Marketing Strategy · E-Commerce Growth · Digital FTE Productization



🗺️  2026 Roadmap

%%{init: {'theme': 'dark'}}%%
mindmap
  root((2026 Focus))
    Agent Protocols
      MCP
      A2A Protocol
      Claude Agent SDK
      OpenAI Agents SDK
    Multi-Agent Systems
      LangGraph
      CrewAI
      AutoGen
      OpenAI Swarm
    Observability
      OpenTelemetry
      Grafana Stack
      Loki + Tempo
    Vector Databases
      Pinecone
      Qdrant
      Chroma
      Weaviate
    Edge AI
      WebAssembly
      ONNX Runtime
      On-device LLMs
    Platform Engineering
      Backstage
      Crossplane
      Terraform
Loading

Area Technologies Why It Matters
Agent Protocols MCP, A2A, Claude Agent SDK Standardizing how AI agents communicate and use tools
Multi-Agent Systems LangGraph, CrewAI, AutoGen, Swarm Orchestrating specialized agents for complex workflows
Observability OpenTelemetry, Grafana (Loki + Tempo) Unified telemetry for AI-native applications
Vector Databases Pinecone, Qdrant, Chroma, Weaviate Scaling RAG systems to production
Edge AI WebAssembly, ONNX Runtime Running inference at the edge without cloud dependency
Platform Engineering Backstage, Crossplane, Terraform Internal developer platforms for AI workloads
AI Safety Constitutional AI, RLHF, HITL Ensuring AI systems are safe and aligned


📊  GitHub Analytics

trophy


GitHub Stats GitHub Streak

Top Languages


Activity Graph


github-snake


🎯  Goals

 Completed

  ✅ Complete all 6 Panaversity Hackathons (Bronze → Platinum)
  ✅ Build cloud-native system with K8s, Kafka, and Dapr
  ✅ Implement Constitutional AI safety with HITL
  ✅ Ship production multi-tenant SaaS (Textile ERP Platform — live on Koyeb + Vercel)

 In Progress

  ⬜ Build multi-agent system with MCP & A2A protocols
  ⬜ Contribute to 3+ open-source AI/ML projects
  ⬜ Publish 24+ technical articles and videos
  ⬜ Launch course on Spec-Driven AI Development
  ⬜ Grow YouTube channel to 1K+ subscribers



💡  Philosophy

"Traditional approach: Avoid AI mistakes. My approach: Learn FROM AI mistakes. Because real innovation happens at the edges of failure."


Principle Practice
01 Spec-First No code without a specification
02 Production Quality Every project is deployment-ready
03 AI as Collaborator Not just a tool — a thinking partner
04 Open Source Share knowledge, elevate the community


✍️  Writing & Content



Agentic AI · Spec-Driven Development · Cloud-Native Architecture · Constitutional AI Safety · Multi-Agent Systems




Let's Build Something Together

Open to collaborating on AI/ML projects, cloud-native systems, and hackathon partnerships.




Profile Views


Build agents that act — not just respond.

Pinned Loading

  1. hackathon-completion-engine hackathon-completion-engine Public

    Cloud-native AI-powered Todo app with Constitutional AI, Kubernetes, Kafka, Dapr, and Discord bot

    Python