AI Engineer specializing in Generative AI, Agentic AI, RAG, and AI Automation, with hands-on experience developing LLM-powered applications, AI agents with tool integration, and intelligent automation pipelines using LangChain, LangGraph, and vector databases. Brings strong Machine Learning and Deep Learning foundations — spanning CNNs, RNNs, transfer learning, and model optimization — applied across real-world computer vision and medical imaging projects to deliver scalable, production-ready AI solutions.
AI Engineering Intern — ATS, UK (Remote) · Apr 2025 – Jun 2025 Designed and built stateful multi-agent workflows in LangGraph, using conditional routing and checkpointed memory to handle multi-step reasoning tasks reliably across sessions. • Standardized how agents connected to external tools by adopting MCP (Model Context Protocol), cutting down on ad-hoc integration code and making tool access consistent across agents. • Added structured output validation and human-in-the-loop approval steps at key decision points — a deliberate tradeoff to keep agents fast without giving up control over high-stakes actions. • Built a RAG pipeline for grounding agent responses in real documents, which noticeably cut down on hallucinated answers compared to plain LLM prompting. • Used LangSmith to trace and debug multi-step agent workflows, quickly identifying where failures occurred.
AI Engineering Intern — Comrex Pvt Ltd · Jul 2024 – Sep 2024 • Built end-to-end RAG pipelines using LangChain, integrating vector databases and document embeddings to deliver context-aware LLM responses. • Developed multi-step reasoning workflows using LangChain agents for complex question-answering and task automation. • Designed and experimented with Agentic AI systems enabling LLMs to autonomously interact with external tools and APIs. • Applied advanced prompt engineering and tuning techniques to enhance contextual understanding and significantly reduce hallucinations. • Integrated Generative AI solutions into production-style applications: chatbots, Q&A systems, and knowledge-based assistants.
AI Intern — Ministry of IT & Telecom, Pakistan · Aug 2023 – Oct 2023 Conducted data preprocessing and exploratory data analysis (EDA) on structured government datasets to extract actionable insights. • Developed Python-based data visualizations using Pandas, NumPy, and Matplotlib to identify trends, patterns, and KPIs. • Applied foundational ML techniques for prediction and classification; gained exposure to government IT infrastructure and digital transformation initiatives. • Optimized ML models via hyperparameter tuning and PCA-based feature engineering, improving accuracy across regression and classification tasks. • Applied transfer learning with pre-trained CNNs (ResNet, DenseNet, EfficientNet, YOLOv9) and LSTM/GRU models, along with data augmentation to boost robustness on limited datasets. • Explored deep learning-based document classification using CNN architectures to automatically categorize scanned government documents, supporting digital record-keeping initiatives.
PyTorch YOLOv9 DenseNet Grad-CAM++ Pydicom OpenCV
A deep learning pipeline for automated spinal pathology detection, combining YOLOv9 for object detection with DenseNet for classification on annotated medical X-ray datasets. Used Pydicom to handle DICOM medical imaging formats and Grad-CAM++ to make model predictions visually explainable — critical for any clinical-adjacent AI system. Applied transfer learning and advanced augmentation to maintain accuracy despite limited medical training data.
LangChain LangGraph Gemini Tavily RAG Agentic AI
A fully agentic research pipeline built from scratch: a Search Agent retrieves live information via the Tavily API, a Reader Agent autonomously selects and scrapes the most relevant source (with PDF-aware extraction, including auto-following arXiv papers to full text), a Writer chain synthesizes a structured report grounded in that retrieved evidence, and a Critic chain self-reviews the output and scores it. Every tool call is logged and traced to verify the agents are genuinely grounding answers in retrieved data rather than hallucinating from memory — a real implementation of retrieval-grounded, multi-agent reasoning with LCEL (LangChain Expression Language) pipelines.
🎙️ AI Video Assistant (https://github.com/imaduddin005)
Whisper · Mistral AI · RAG · LangSmith · Streamlit · GitHub Built an end-to-end pipeline that transcribes YouTube videos or local audio/video files using Whisper and runs summarization, action-item, and key-decision extraction in parallel using Mistral AI. • Implemented a RAG-based chatbot over each transcript using a Chroma vector store, rebuilt per session to keep answers grounded strictly in the correct video’s content. • Integrated LangSmith tracing to monitor prompts, token usage, and latency, and built both a Streamlit web interface and a command-line version for running the pipeline.
Python · Streamlit · BeautifulSoup · SQLite · Ollama (Gemma 3) · GitHub Built an autonomous discovery agent that finds software houses and tech companies across major Pakistani cities that genuinely work with AI, using automated web search, scraping, and deterministic keyword classification. • Designed a hybrid classification pipeline: fast rule-based keyword matching handles most companies, with a cached, local LLM (Ollama + Gemma 3) call used only for ambiguous cases to minimize token cost. • Implemented SQLite-backed caching for scraped domains and LLM classifications, avoiding redundant network requests and re-classification on repeated searches. • Shipped both a Streamlit web interface and a CLI, letting users filter results by city and AI domain (Machine Learning, GenAI, Agentic AI, RAG, NLP, etc.).
Languages: Python · C++
GenAI & LLM: LangChain · LangGraph · MCP (Model Context Protocol) · RAG Pipelines · Prompt Engineering · Vector Databases
AI Automation n8n . Claude Code . Open AI Api . Workflow Automation . Agentic Ai pipline
ML / DL Frameworks: PyTorch · TensorFlow · Keras · Scikit-learn
Computer Vision: YOLOv9 · ResNet · DenseNet · EfficientNet · OpenCV · Transfer Learning · Grad-CAM++
Data & Visualization: Pandas · NumPy · Matplotlib · Feature Engineering · PCA · Data Augmentation
Backend & APIs: FastAPI · Flask · REST APIs
Databases: MySQL · PostgreSQL
Tools: Git · GitHub · Jupyter Notebook · Google Colab · Kaggle · VS Code
Methods: Supervised & Unsupervised Learning · Hyperparameter Tuning · LSTM/GRU · Transformer Models
Open to AI Engineering roles, research collaboration, and challenging agentic AI / RAG problems.