Muhammad Haris Altaf
Computer Science student at CUST (2024–2028) specializing in AI, Agentic Systems, and Full-Stack development. Independently built 7 production-style AI projects including a multi-agent LLM pipeline, two RAG systems with hybrid vector search, and a voice-controlled automation assistant. Proficient in Python, FastAPI, LangChain-style agentic workflows, LLM integration, and Next.js. Actively seeking internship opportunities in AI engineering and applied machine learning.
- Spent a year getting solid hands-on experience with relational databases, writing SQL queries, and understanding how data is structured and managed.
- Covered advanced DBMS concepts through university and applied them in real project work including data modeling and normalization.
- Worked with clients worldwide handling large-scale data entry, pulling business information from websites, generating leads, and conducting web research.
- Comfortable working with MS Excel and Google Sheets, collecting structured data at scale around 1000+ records with clean accuracy.
Designed and built a 4-agent pipeline (Planner, Researcher, Writer, Critic) that breaks down complex goals into sub-tasks, executes them in sequence, and delivers output in three formats (text, structured JSON, HTML report) with real-time execution logging.
Built a full-stack RAG system supporting any uploaded document with hybrid retrieval (vector + keyword search) and streaming LLM responses. Features a custom-designed dark UI and sub-2-second query response time.
Developed a production-ready RAG platform with TF-IDF + BM25 fusion ranking and multi-format document ingestion (PDF, DOCX, CSV, MD). Supports API key-based access and streaming LLM responses.
Built a voice-activated AI assistant capable of sending WhatsApp messages, performing web searches, and managing desktop files entirely through natural speech commands with no manual input required.
Developed a full-stack AI avatar trained on a personal persona, deployable as a FastAPI backend or standalone web app. Allows users to interact with a conversational clone via a browser interface.
Built an automated news aggregator that fetches, filters, and summarises daily news digests via NewsAPI. Delivered through a production-grade web interface with zero backend required on the client side.
Developed an AI agent that classifies incoming emails, auto-replies to routine queries, and flags high-priority messages (e.g. payment-related) for human review, reducing manual email handling time.