🤖 AI/ML Engineer building production agentic AI and voice AI systems 📊 Data Scientist — Python, SQL, Power BI & applied ML/DL

💼 AI Engineer in production voice/GenAI systems — building real-time voice agents, RAG pipelines & multi-tenant AI platforms ⭐ ML/Data Science Freelancer @ Fiverr — 100% Response Rate 🎓 BS Computer Science @ Superior University, Lahore (2022–2026) 🏆 Kaggle Notebooks Expert | Google Data Analytics Certified 🚀 Currently building: Urdu-first voice AI for healthcare, multi-tenant LiveKit ordering pipelines, and RAG/agentic system

C++ games & Android apps  →  Web dev (Flask, React Native)  →  Data Science & EDA  →  Agentic AI & Voice AI
      (2023)                       (2023–2024)                    (2024)                (2025–2026)

Started with a bouncing-ball game and a chess engine in C++ during my first two semesters, moved into web development, then spent a year deep in data science — EDA, Kaggle notebooks, classical ML. Since 2025 I've been building production agentic and voice AI systems: RAG chatbots, MCP servers, and multi-tenant voice agents used in real hospital and restaurant workflows. Every stage is still up here — I'd rather show the full arc than hide where I started.

Final Year Project (Superior University, supervised by Dr. Jawad Ahmad, built with a 2-person team). A realtime, Urdu-first voice AI agent that lets patients book, reschedule, and cancel hospital appointments over a live phone call — multi-tenant SaaS architecture with RBAC, call logging, and an analytics dashboard. Tech: Python · FastAPI · VAPI · GPT-4o (tool calling) · LangChain · LangGraph · Redis · Twilio · PostgreSQL 🔗 Live demo · AI Agent repo · Frontend repo

End-to-end ML pipeline for Lahore property data: scraping → cleaning → feature engineering → 5 regression models compared (Linear, Ridge, Lasso, Random Forest, XGBoost) with cross-validation and hyperparameter tuning, a TF-IDF + cosine-similarity content-based property recommender, and an Explainable AI module, shipped as a multi-page Streamlit app. Tech: Python · scikit-learn · XGBoost · Pandas · Streamlit 🔗 Repo

• YouTube RAG Chatbot — Full RAG pipeline over YouTube transcripts: chunking, 384-dim sentence-transformer embeddings, Pinecone vector DB with custom namespaces, Groq LLM, LangChain orchestration, Streamlit UI with session memory.
• Cold Email Generator — LLaMA-orchestrated pipeline that scrapes a company's careers page, extracts role requirements, vector-matches them against portfolio projects, and drafts a personalized cold email.
• n8n Personal Assistant — Natural-language agent orchestrating Google Calendar, Gmail, Tasks, Docs, and Sheets through n8n workflow automation.
• VoxKit Voice Agent (LiveKit) — Modular-monolith, multi-tenant voice ordering system for restaurants, built on LiveKit + FastAPI + GPT-4o. (production work — link to the relevant repo once confirmed public/shareable.)
• Hybrid Voice Pipeline — OpenAI STT+VAD → LangChain → ElevenLabs TTS over Twilio; ~15x cheaper than the OpenAI Realtime API while preserving barge-in and function calling.

• PSL Analytics Platform — 73,785 rows of ball-by-ball PSL cricket data behind a public REST API (FastAPI, deployed on Vercel) plus a Streamlit dashboard; backed by two self-published Kaggle datasets.
• SMS Spam Classifier — TF-IDF + Naive Bayes + stacking ensembles on 5,572 SMS messages: 100% precision, 95–97% accuracy, deployed with Streamlit.
• Cats vs Dogs CNN — TensorFlow/Keras CNN with a full optimization journey (baseline → dropout → batch norm → L2): 81.85% validation accuracy, 88.71% train accuracy.
• WhatsApp Chat Analyzer — Streamlit app for message stats, timelines, word clouds, activity heatmaps, and emoji analysis.
• Interactive Power BI Dashboards — HR attrition analytics, e-commerce sales, hotel-cancellation analysis, and a 10-year non-profit financial EDA across 3 charities.

• Stepstone Job Market Analyzer — Selenium scraper covering 1,300+ Düsseldorf tech job listings, handling pagination, cookies, and dynamic scroll.
• Yahoo Finance Scraper — OOP-refactored Selenium scraper pulling real-time price, volume, market cap, and P/E data for 50+ stocks daily.