Zeeshan Zafar

Zeeshan Zafar Islamabad,Pakistan Ó - - Linked-In ‡ Zeeshan0123 EDUCATION National University of Computer and Emerging Science(FAST) Bachelor Of Computer Science à Portfolio August 2020 – May 2024 Islamabad, Pakistan EXPERIENCE ResouceX.IO AI Engineer • • • • • • • • 2025 present Built an end-to-end AI pipeline, including fine-tuning a Large Language Model (LLM) tailored for the repair industry using an in-house 1.6 million record dataset. Designed an advanced Agentic RAG pipeline leveraging Qdrant for vector storage, an advanced reranker for improved retrieval accuracy, and web search agents for real-time data integration. Deployed AI pipelines and solutions on Azure Virtual Machines, ensuring scalability and reliability in production environments. Managed large datasets across multiple collections, ensuring high performance and efficient processing. Optimized workflows using Hugging Face embeddings tailored to specific use cases and integrated Open and Close Source LLMs where appropriate. Generated synthetic datasets from unstructured data to enhance model training and accuracy. Developed and deployed AI-driven features seamlessly integrated into the company’s customer-facing portal. Collaborated with 4+ AI engineers to improve the performance of AI pipelines. CafeVista Pvt Ltd AI Engineer • • • • • • • 2024 past Fine-tuned diffusion models for text-to-image and image-to-text applications, enhancing model accuracy and output quality Designed and implemented FastAPI backends for mobile and web applications, ensuring seamless integration and high performance. Built interactive AI-powered user interfaces using Next JS, Gradio, Chainlit, and Streamlit, enhancing user engagement. Created and deployed Retrieval-Augmented Generation (RAG) pipelines with LangChain and LlamaIndex, leveraging AWS and Hugging Face for scalability. Deployed APIs and applications on Hugging Face cloud, ensuring robust and scalable AI solutions. Implemented scalable AI workflows on AWS, optimizing performance and integration in production environments. Collaborated cross-functionally to deliver AI-driven solutions, improving operational efficiency and user engagement. PROJECTS Antobot: Anatomy Chatbot • • • • • • • • | Groq (Llama-4-Scout), PipeCat, Distil-Whisper, PlayHT Groq (Llama-4-Scout), PipeCat, Distil-Whisper, PlayHT Function calling for dynamic API integrations.. Deployable via PipeCat Cloud (+ Twilio telephony support). AI Company Lead Gen Bot • LlamaIndex,LORA/QLora,LlamaCpp ,Bio-Mistral,Chainlit UI, RAG AI-powered chatbot for medical students to learn anatomy through interactive conversations 3D body part visualization. Quiz generation system. Personalized learning paths. Real-Time Voice AI Agent • | | Ollama,Selenium,Langchain,Transformers,Huggingface,Next Js Automated lead generation system using web scraping from Linked In, automation, and a RAG-based chat app. LinkedIn data scraping. Automated outreach and follow-ups. • • RAG-based bot for personalized interactions. Integration using Next Js. AI-Powered Storytelling Bot • • • • | Gradio,Diffusion,Langchain,FluxAI,FastAPI,Hugging Face Spaces,Docker A production-ready API that generates structured stories with dynamic illustrations using advanced AI models Custom story generation based on mood, theme, and age. AI-generated illustrations for different scenes. Real-time API responses with scalable backend Gradio App for UI and deployment. TECHNICAL SKILLS No-Code-Tools: n8n, Vapi(voice AI Agents), Bolt, Lovable Languages: Python, C, C++, Dart, JavaScript,Type Script, SQL, NoSQL, React AI/ML:TensorFlow,PyTorch,Huggingface,Scikitlearn,Langchain,Llamaindex,Diffusion,Transformers,OpenCV, MLflow Tools and PlatformsAWS (SageMaker, Lambda, Bedrock, EC2),Azure (ML Services, virtual machine),GCP (Vertex AI),Groq,Docker,Git,Github,FastAPI,Flask,Next JS,Unsloth(finetuning),Llama Cpp,Ollama,Crawl4AI Agentic Frameworks Pydantic AI, LangGraph Vector DBs Croma,Qdrant,FAISS,Weaviate CERTIFICATIONS • • Deep Learning Specialization - Coursera Natural Language Processing - Stanford Online • Advanced Machine Learning - Google Technical Achievements • • 90% accuracy on MMLU Anatomy dataset of my own model Antobot • Built production-ready RAG systems Publish my Own Models on HuggingFace Hub