Available for AI projects
I build production-ready AI agents, RAG systems, and intelligent backend workflows.
Working with Python, FastAPI, LangChain, OpenAI, Claude, Ollama, vector databases, and modern deployment workflows to turn AI ideas into reliable products.
I am Muhammad Kashif, an AI Engineer focused on building LLM applications, agentic workflows, RAG systems, and backend services that turn complex ideas into useful products.
My approach starts with the problem, not the model. I care about clean APIs, structured outputs, strong retrieval quality, thoughtful prompts, and systems that can be tested, monitored, and improved over time.
I began with Python, data processing, and machine learning fundamentals, then moved deeper into LLM-powered applications using FastAPI, LangChain, OpenAI, Claude, Ollama, FAISS, Pinecone, and ChromaDB.
FAST National University, Islamabad. Built a foundation in AI, programming, data, and intelligent systems.
Strengthened Python, OOP, data structures, scripting, NumPy, Pandas, Matplotlib, and clean coding practices.
Built LLM workflows, RAG systems, FastAPI services, REST APIs, and deployed scalable AI solutions with Docker, AWS, Git, and CI/CD.
Designing intelligent workflows that combine reasoning, tools, APIs, and automation.
Building data-driven models, pipelines, evaluation flows, and practical ML solutions.
Working with neural networks for classification, NLP, and vision-focused systems.
Creating image classification and detection systems for real-world visual tasks.
Extracting meaning from text with parsing, summarization, sentiment, and intent flows.
Developing LLM apps with prompt engineering, RAG, tool calling, and structured outputs.
Writing clean Python for AI services, data workflows, APIs, automation, and scripting.
Designing production APIs, microservices, integrations, and AI-ready backend systems.
Building clean interfaces and portfolio experiences with responsive HTML, CSS, and JS.
Deploying and connecting AI applications with cloud databases, hosting, and storage.
Using Docker, Git, CI/CD, monitoring, logging, and repeatable deployment workflows.
A modern timeline of roles, achievements, and technologies across AI engineering, backend services, Python development, and mobile AI integration.
Applied research direction: LLM apps, NLP workflows, and Flask/FastAPI services.
ViewPython workflows for data science, analysis, and AI foundations.
ViewEvery project is framed as a product story: the problem, system design, implementation tradeoffs, and the result it was built to create.
Local knowledge-base chatbot that retrieves context, reasons over documents, and generates grounded answers with tool-ready workflows.
Generic chatbots cannot answer reliably from private documents without retrieval and grounding.
Built ingestion, chunking, embeddings, semantic search, prompt orchestration, and response generation.
Documents -> chunks -> embeddings -> FAISS index -> retriever -> LLM answer pipeline.
Improving chunk quality, retrieval relevance, and context-aware prompts for better answers.
A reusable RAG foundation for custom Q&A over private knowledge bases.
Automated document intelligence pipeline for extracting, chunking, summarizing, organizing, and reusing document knowledge.
Manual document review is slow and makes stored knowledge hard to search or reuse.
Created APIs for parsing, chunking, summarization, storage, and searchable document intelligence.
Upload -> parse -> chunk -> summarize with Ollama -> store in Supabase -> searchable API.
Keeping extraction repeatable while handling different document structures and summaries.
A maintainable automation pipeline for reusable document workflows.
An AI-powered emotional digital twin concept using intent detection, sentiment analysis, LLM responses, and speech interaction.
Emotion-aware conversations require more than generic text generation.
Combined sentiment, intent, prompt orchestration, STT/TTS, and LLM-based response generation.
User input -> intent/sentiment layer -> LLM response -> voice output through audio APIs.
Balancing natural tone, context, latency, and emotionally appropriate outputs.
A prototype direction for more natural human-AI emotional interaction.
Real-time computer vision system for detecting and classifying textile defects to support automated quality inspection.
Manual textile quality inspection is repetitive and inconsistent at production speed.
Trained and tested a YOLO-based detection pipeline for visual fabric defects.
Image/video input -> preprocessing -> YOLOv8 inference -> defect labels and bounding boxes.
Dataset preparation, model evaluation, and reducing false detections across defect types.
A practical inspection concept to reduce manual checking effort in textile workflows.
TypeScript web application deployed on Vercel, positioned as an AI-powered product experience.
AI product ideas need a clean, accessible web interface to become usable.
Built a TypeScript app with a deployment-ready frontend hosted on Vercel.
Client application -> product UI -> hosted deployment pipeline on Vercel.
Packaging the experience into a polished, shareable interface.
A live deployed project users can open and evaluate directly.
End-to-end MLOps project using federated learning for privacy-aware health risk prediction.
Health prediction systems need model workflows while respecting distributed sensitive data.
Explored federated learning and MLOps patterns for health-risk prediction.
Distributed training concept -> model workflow -> health risk output pipeline.
Designing ML workflow automation around privacy-aware model development.
A foundation for scalable healthcare ML experimentation.
Course activity automating an ML workflow with preprocessing, training, evaluation, containerization, and GitHub Actions.
ML projects often fail to automate training, validation, and packaging repeatably.
Built a GitHub Actions workflow around the ML lifecycle.
Preprocess -> train -> evaluate -> containerize -> CI/CD workflow.
Keeping pipeline steps reproducible and automation-friendly.
A compact reference workflow for ML automation practice.
Personal portfolio website rebuilt into a premium AI Engineer showcase with strong storytelling and interactive sections.
The portfolio needed stronger positioning for AI engineering and premium presentation.
Built a modern design system, navigation, hero, story, skills, and case-study showcase.
Static frontend assets served by Express with deploy-ready configuration.
Improving polish while preserving section-by-section scope.
A cleaner personal brand surface for AI, LLM, RAG, and backend work.
Collection of Flutter/Dart app work connected to mobile UI, reports, Firebase, and app development practice.
Mobile AI/product workflows need usable app interfaces and reliable data flow.
Built Flutter/Dart app experiments around mobile screens and backend-connected workflows.
Flutter UI -> app state -> Firebase/backend integrations.
Keeping UI responsive while integrating app data and services.
A repository of mobile development practice and project work.
No projects match your search yet.
Have an LLM app, RAG workflow, AI automation, or backend system in mind? Send a short brief and I’ll get back with a practical next step.