Hafiz Arham
AI Engineer specializing in Agentic AI, LLM Applications, and Deep Learning.
Based in: Lahore, PakistanCurrently: Junior Lecturer, Object-Oriented Programming, Superior UniversityExperience: 1 year
An AI Engineer who builds end-to-end LLM agents and deep learning models to automate complex business operations like lead handling and bookings. With a background in full-stack and Android development, he delivers complete AI products from multi-agent workflows to medical imaging ensembles.
Experience
Junior Lecturer, Object-Oriented Programming
Superior UniversityLahoreworkSep 2026 – Present
Teach OOP to undergraduates through lectures and hands-on labs; design programming tasks and assessments.
Object-Oriented Programming
AI Automation Engineer
Roomello (UK)RemoteworkAug 2026 – Present
Summary
Automate support, lead handling and bookings for a UK short-stay business: AI replies to DMs and post comments on WhatsApp, Instagram, Facebook and TikTok, plus website and email leads.
What I did
- Built an integration to automatically assign tasks from incoming leads.
- Implemented multi-agent handoffs using CrewAI to split lead-to-booking flows into specialized agents for intake, pricing, and execution.
- Used Model Context Protocol to standardize LLM interfaces with inventory databases and calendar tools via context servers and resource endpoints.
- Enforced strict structured outputs using Pydantic models and tool-calling frameworks to ensure LLM responses adhere to precise JSON schemas.
- Orchestrated multiple external APIs including inventory/availability, pricing/dynamic discount engines, and payment gateways for property reservations.
- Implemented a tool-calling wrapper that catches specific error codes and feeds them back into the model's context as structured observations to prompt corrective actions.
Results
- Ensured graceful degradation and maintained system state integrity by managing exceptions through defined recovery paths when unexpected API results occurred.
WhatsAppInstagramFacebookTikTokEmailCrewAIModel Context Protocol (MCP)OpenAI Structured OutputsInstructorJSON Schema
AI Automation Engineer
Signature LabelsLahoreworkMar 2026 – Jul 2026
Summary
Built a WhatsApp AI concierge for a luxury car hire business (Next.js, Vercel AI SDK, LLM tool calling): availability, live pricing, reservations and post-rental support; replaced an older n8n workflow. Built a WhatsApp booking agent in n8n for a plumbing firm: qualifies leads, sends quotes, dispatches plumbers.
What I did
- Automated support and lead tracking using a WhatsApp system.
- Orchestrated multiple external APIs in omnichannel booking and customer support workflows using robust error handling and fallback logic.
Next.jsVercel AI SDKLLM tool callingn8nWhatsApp
Skills
Technical
Prompt Engineering
PythonMulti-agent systems
WhatsApp Business API
n8n
Agentic AI
LLM tool calling
Computer Vision
GitRAG
LangChain
PyTorch
TensorFlowCNNs
NLP
Meta Graph API
Webhooks
REST APIs
JavaScript
TypeScriptJava
Next.js
ReactAndroid (Java / XML)
CrewAI
FastAPIElevenLabs
Node.jsGANs
scikit-learn
Vision Transformers
MCP
Twilio
Pydantic
Tool use orchestration
Autonomous Planning
Reflection loops
Fréchet Inception Distance
Visual Turing Test
ReAct (Reason + Act)Model Context Protocol (MCP)
Projects
Multi-Class Diabetic Retinopathy Detection
Lead Developeracademic
Five-level severity grading of retinal fundus images. Fine-tuned EfficientNet (CNN) and Swin Transformer (ViT), then combined them with a weighted soft-voting ensemble that beat both single models.
EfficientNetCNNSwin TransformerViTSoft-voting ensemble
GAN-Based Realistic Human Signature Generation
AI Researcherpersonal
Trained a GAN to generate realistic handwritten signatures, from dataset preparation to adversarial training.
GANGenerative AIModel training
AI Voice Calling Agent
Developerpersonal
Handles phone calls, books meetings and emails follow-ups, with a React dashboard for call logs.
Agentic AIReactVoice AI
Full-Stack & Mobile Development
Full-Stack Developerpersonal
Service website: built end to end (front end, back-end APIs, database, deployment). PrepXpert: Android app (Java / XML) that turns a student's notes or syllabus into AI-generated study notes and MCQ tests.
JavaXMLAndroidAPIsDatabase
Omnichannel AI Lead-to-Booking Agent
AI Engineerprofessional
One LLM agent behind every channel answers questions on properties, dates and prices, qualifies the lead and books it through to payment, greatly reducing the lead handler's manual work.
LLMWhatsAppInstagramFacebookTikTokReAct (Reason + Act) loop
Education
BS Computer Science
Superior UniversityComputer ScienceGPA 3.632022 – 2026
Thesis. Multi-Class Diabetic Retinopathy Detection
Recognition
Certifications
Claude with the Anthropic API
Anthropic
Claude Code
Anthropic
Model Context Protocol (Intro & Advanced)
Anthropic
Multi-Agent Systems
CrewAI
Fundamentals of Generative AI
Microsoft
n8n AI Agent Builder
Simplilearn
Hafiz's twin is AI, it can make mistakes.
More people on Kanvis
Aleena masood
Lahore
Artificial Intelligence Student & Software Intern specializing in RAG pipelines and Generative AI.
Urdu · English · LLM Integration · RAG
kanvis.me/aleena-masood
Bilal Ahmed
Senior AI Engineer specializing in Agentic AI, LLM Orchestration, and Real-Time Voice Systems.
Python · English · Urdu · DSP
kanvis.me/bilal-ahmed1
Muhammad Noman
Dammam
Python Backend & AI Engineer specializing in Scalable Systems and Generative AI
Urdu · Django REST Framework (DRF) · Django · Python
kanvis.me/muhammad-noman
abdul hadi
Karachi
AI Engineer specializing in multi-agent orchestration and high-performance model optimization.
Python · Ollama · Multi-agent orchestration · REST APIs
kanvis.me/abdul-hadi2
