BS Artificial Intelligence @ GIKI · CGPA 3.65 · Pakistan

AI Engineer focused on building production-grade intelligent systems — multi-agent LLM pipelines, ML platforms, MLOps infrastructure, and generative models. Three remote AI/ML internships completed. I build things meant to actually run.

My work sits at the intersection of AI agents, data systems, and deployment infrastructure. I care less about notebooks and more about systems that hold up under real conditions.

Flagship projects:

• EconML — Full-stack financial ML platform. 26 models across 8 assets (AAPL, MSFT, GOOGL, NVDA, TSLA, BTC, ETH, SOL) covering regression (XGBoost, R²=0.78), BiLSTM forecasting (6.18% MAPE), KMeans clustering, and SMOTE-based fraud detection (98% AUC-ROC). Stack: FastAPI, Docker, Prefect, GitHub Actions CI/CD.

• The AI Scientist — Multi-agent LLM research assistant built for the World Bank × Hack-Nation AI Hackathon. Three coordinated sub-agents (Literature Scout, Hypothesis Generator, Experiment Planner) retrieve live data from ArXiv and Semantic Scholar. Runs on LLaMA 3.3 70B via Groq and Gemini 2.0.

• Agentic Multimodal RAG — Cross-modal retrieval pipeline combining CLIP embeddings, FAISS HNSW indexing, Neo4j knowledge graph, LangChain ReAct agent, and BLIP-2/LLaVA for visual question answering.

• Generative Fashion Designer — Generative suite for textile pattern design. Six locally-trained models: VAE, DCGAN, WGAN-GP, cGAN, Latent DiT, and Flux API. Trained on DTD dataset (5,640 images, 47 classes). Quantitative evaluation via FID, IS, and SSIM. Containerized REST API with React frontend and GitHub Actions CI/CD.

• Aegis-IoT — 5-layer AnyLogic IoT network simulation with embedded ML inference (Random Forest, One-Class SVM, BiLSTM transpiled via m2cgen), tabular Q-learning for adaptive congestion control, and FedAvg federated learning across 3 edge nodes. Uses RT-IoT2022 dataset (4,132 flows) with GAN-style DDoS stress injection. React + NOC dashboard.

AI / ML PyTorch · TensorFlow · Scikit-learn · XGBoost · Hugging Face Transformers · LangChain · Sentence-BERT · BART · CLIP · BLIP-2 · LLaVA · Stable Diffusion · Optuna · SMOTE

Agents & RAG LangChain ReAct · Multi-agent orchestration · FAISS HNSW · Neo4j · Vector DBs · Retrieval pipelines · ArXiv / Semantic Scholar APIs

MLOps & DevOps Docker · Docker Compose · GitHub Actions · Prefect · FastAPI · Terraform · AWS EC2 / ECR / ECS · Bash scripting

Languages & Data Python · C++ · JavaScript · Bash · Pandas · NumPy · Gradio · Flask · MongoDB · PostgreSQL

NLP Intern — Elvvo Built production NLP pipelines using Sentence-BERT for semantic resume screening and BART for multi-document summarization. Worked on real hiring data with production-level constraints.

AI/ML Intern — CodeAlpha Developed deep learning models and data preprocessing pipelines. Built handwritten character recognition system (TensorFlow CNN on EMNIST).

AI/ML Intern — Developers Hub Corporation Worked on scalable ML solutions, data pipeline integration, and AI agent implementation.

BS Artificial Intelligence — Ghulam Ishaq Khan Institute (GIKI) 2023 – 2027 · CGPA: 3.62 · Upward trajectory across all semesters

Relevant coursework: Deep Neural Networks, Machine Learning, Computer Vision, Data Structures & Algorithms, Computer Communications & Networks, DevOps, Big Data Analytics, Knowledge Representation

Email: zainulabdeen9909@gmail.com LinkedIn: linkedin.com/in/zain-ul-abdeen-48aa72318 Portfolio: https://zain-ul-abdeen-773.netlify.app/)