Who is Shaheer?
Introduction, biography and personal manifest.
I'm Shaheer, a final-year AI student who's been building actual systems for the past two years. MLOps pipelines, RAG architectures, backend APIs, computer vision apps. Mostly at companies that needed them to work. I'm not the type to stop at getting something running. I'll get it working, then go read the paper behind it to understand why. That habit has made me a better engineer than any tutorial did. I've done my best work in the deep end in unfamiliar frameworks, Hackathons helped me, so did shipping things and having to fix them. What I'm looking for: a team working on problems hard enough that the answer isn't obvious. Somewhere that cares about the work, not just the output.
Based in Karachi. Open to on-site and remote roles in AI/ML engineering and backend
Career changelog and roles.
10Pearls
Resolve (Research Solutions and Ventures)
Arcana Info Pvt Ltd
Index of 12 repositories. Hover to preview details.
Visual Product Retrieval. Content-based image retrieval system using MobileNetV2 for feature extraction, PCA for dimensionality reduction, and FAISS for similarity search. Includes K-Means clustering and Linear SVM classification for product categorization.
Academic Paper Assistant. Intelligent research tool combining Gemini AI with Model Context Protocol servers (arXiv, web fetch, filesystem). Organizes papers by topic in structured JSON, provides natural language search, and includes pre-built research prompts.
AI Food Recognition. Computer vision app identifying Pakistani foods with custom CNN (PyTorch). Deployed on Azure ML with REST API, MongoDB for nutrition data, and Streamlit frontend. 80%+ accuracy across 6 food categories.
Gesture-Controlled Chess. Computer vision chess game using MediaPipe hand tracking for touchless piece control. Features full chess logic including castling and en passant, AI opponent, and overlay interface combining webcam feed with game board.
AI Code Analysis Tool. Built with React and TypeScript, combining Gemini 2.0 AI with static code analysis to instantly explain any GitHub repository. Generates architecture diagrams using Mermaid.js, provides contextual chat, and features interactive file exploration with smart caching.
AI Resume Tailoring. Next.js app that uses Gemini AI to auto-tailor resumes to job descriptions. Integrates GitHub API to select relevant projects, generates match scores, and exports ATS-friendly PDFs. Privacy-focused with browser-only storage.
RAG Document Assistant. Q&A system for Pakistan's federal budget using RAG architecture. Processes 300+ page documents with LangChain and LLaMA-3-8B, achieving 85%+ answer accuracy.
MLOps Pipeline. End-to-end ML pipeline with automated retraining using GitHub Actions and Hopsworks. XGBoost model with batch inference and continuous updates from user data.
Multi-Store Backend. Scalable inventory API evolving from SQLite to PostgreSQL with Redis caching, WebSocket sync, API authentication, and RBAC. Docker-ready with comprehensive audit logging.
LinkedIn Prompt Decoder. Streamlit app that reverse-engineers AI-generated LinkedIn posts to reveal the original prompts. Uses Gemini API for pattern recognition and prompt reconstruction.
Real-Time Fact Checker. Browser extension (Firefox + Chrome) with context menu integration for instant fact-checking. Right-click any text to verify claims using Gemini API. Manifest V2/V3 compatible.
Sentiment Analysis. LSTM-based sentiment classifier deployed as a web app. Trained on text data to predict emotional tone, with saved tokenizer and label encoder for consistent inference.
Academic background and degrees.
Bachelor of Science in Artificial Intelligence
Certifications and licenses.