Waqar
Experience
Technical Team Lead
Designed and led the development of the company’s web presence, including familyrealtorspk.com, and an agentic multi-tenant CRM platform with role-based access control and Meta/Google Ads API integration. Oversee planning and improvement of internal software systems, IT operations, data workflows, and DevOps tasks, working closely with the team to build practical technology solutions that support daily real estate operations end-to-end.
Private Tutor
- Individualized tutoring in CCNA networking, data science fundamentals, and full-stack web development
- Windows Server administration including Active Directory, DNS, DHCP, and Group Policy
- Advanced mathematics including calculus and linear algebra
- Customized learning plans and practical exercises to help students understand technical concepts and problem-solving methods
Machine Learning Engineer
- Developed and deployed production ML applications for clients across content generation, real estate, and data extraction domains
- Built agentic AI systems using LangGraph
- Engineered full-stack solutions with Django and FARM stack
- Specialized in RAG systems, LLM fine-tuning, and end-to-end production deployments
Machine Learning Engineer
Completed production projects for clients globally, including a chess CNN model with minimax & alpha-beta pruning to play at professional level, deep learning for live TV ad detection and categorization, and DBSCAN-based sports odds prediction to flag bookmaker discrepancies. Open-source work includes a MCTS-UCB-Transformer ensemble chess engine, BERT fine-tuning for sentiment analysis and essay grading, Llama 2 fine-tuning for physics Q&A, and a decoder-only transformer for generative AI.
Skills
Ⅰ︎ ML & Deep Learning
PyTorch, JAX, Scikit-learn, Pandas, NumPy for end-to-end model development. OpenCV, YOLO, VGG16, CNNs and Vision Transformers for computer vision pipelines.
Ⅱ︎ NLP & LLM Engineering
Fine-tuning Flan-T5, Qwen, and LLaMA series on custom datasets. Production integration with OpenAI, Gemini, Claude, and Hugging Face APIs; agentic workflows with LangGraph.
Ⅲ︎ Full Stack Web Development
FARM stack, Django, React, Next.js - building scalable web applications and REST APIs, including real estate platforms, CRMs, and AI-powered tools.
Ⅳ︎ Cloud & DevOps
Azure, AWS, Railway, Koyeb for serverless and containerized deployments. Docker and Kubernetes for orchestration and ML model hosting at scale.
Ⅴ︎ Data & Databases
Matplotlib for analysis and reporting. PostgreSQL and MongoDB for relational and document-based storage; data pipelines with Pandas and NumPy.
Ⅵ︎ Networking & Systems
CCNA networking fundamentals, Windows Server administration including Active Directory, DNS, DHCP, and Group Policy; IT infrastructure and operations.
Tools & Technologies
Featured Projects
LinkedIn Content Generator Web App
A monolithic progressive web app in Django with PostgreSQL, a complete agentic AI content generator using LangGraph, scheduling and notification via cronjobs and Resend API, Google OAuth, Cloudinary media storage, and Cloudflare security.
Agentic RAG System with Hybrid Search
Production-ready agentic RAG using FARM stack + LangGraph query routing. Self-hosted Modal.com infrastructure serving Qwen 8B on T4 GPUs, BGE-Large embeddings and BGE-Reranker on L4 GPUs. Google OAuth, user rate limiting - containerized for Railway and Koyeb.
Agentic CRM for Real Estate Company
FARM stack CRM with role-based access control, Gemini API, LangGraph, Cloudinary, MongoDB Atlas, WhatsApp and email notifications - fully agentic client management.
Real Estate Web-app
Built with Next.js and MongoDB - property listings, blogs, user submissions with admin approval, full CRUD operations, and a separate admin panel.
LinkedIn Scraper API
A REST API that extracts LinkedIn profile data in real-time using persistent Chrome sessions with one-time authentication to bypass anti-bot measures, delivering structured JSON via HTTP.
Construction Site Portfolio
A dynamic Next.js website for a construction company - WhatsApp API integration, MongoDB database, and a modern content-driven design.
Detecting Ads on Live TV Using Deep Learning
Deep learning architectures to detect commercial starts and categorize ads in live TV broadcasts. Detailed pipeline available - source code under NDA.
Predicting Sports Odds via Unsupervised Learning
DBSCAN-based clustering to find relative average predictions vs. bookmakers; detects errors when variance exceeds threshold for sports betting analysis.
MCTS-UCB-Transformer Ensemble Chess Engine
Chess engine combining an encoder-only Transformer, UCB exploration, and Monte Carlo Tree Search in an ensemble - real-time play via terminal.
Certifications
Let's Build Together
Ready to collaborate on your next project? Reach out