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.
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) |
|
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
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. |
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 "⚙️ Core Platform"
B["⚡ FastAPI Backend<br/><sub>+ Dapr Sidecar</sub>"]
I["🛡️ Constitutional AI<br/><sub>Safety Middleware</sub>"]
end
subgraph "💾 Data Layer"
D["🗄️ PostgreSQL 15<br/><sub>StatefulSet</sub>"]
H["📦 Redis 7<br/><sub>State Store</sub>"]
end
subgraph "📡 Event Streaming"
E["📡 Apache Kafka<br/><sub>Strimzi KRaft</sub>"]
F["🔔 Notification<br/><sub>Service</sub>"]
end
subgraph "📊 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
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 |
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) |
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.
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 productionWhat 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 |
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)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)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
| 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 |
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"
};Textile & CMT Manufacturing · Dubai Real Estate Marketing · Digital Marketing Strategy · E-Commerce Growth · Digital FTE Productization
%%{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
| 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 |
|
✅ Complete all 6 Panaversity Hackathons (Bronze → Platinum)
|
⬜ Build multi-agent system with MCP & A2A protocols
|
"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 |
Agentic AI · Spec-Driven Development · Cloud-Native Architecture · Constitutional AI Safety · Multi-Agent Systems
Open to collaborating on AI/ML projects, cloud-native systems, and hackathon partnerships.

