Pakistan
Hello 👋 I'm Saad, a senior at Habib University with a passion for building intelligent web applications. My expertise combines full stack development with a strong focus on Agentic AI and Large Language Models. I'm currently contributing to open-source projects as a Google DeepMind GSoC participant, enhancing Gemini API integrations in LangChain, Llama-Index, and Agno. I thrive on creating seamless, visually stunning, and AI-powered web solutions. Let's connect to innovate and create together!
◦ Architected an event-driven, async-first Go backend using Redis Streams to process high-throughput traffic for Chatly (3M+ users) and ImagineArt (5M+ users), reducing system errors by 90%. ◦ Engineered real-time AI capabilities, including RAG-based persistent memory, AI Sheets/Docs parsing, and multilingual TTS/STT, by integrating state-of-the-art models. ◦ Optimized PostgreSQL architecture and inverted indexes for rapid, paginated chat retrieval, and developed robust quota middleware to ensure accurate, low-latency billing at consumer scale.
• Led enhancements to Gemini API integrations in LangChain and LlamaIndex, optimizing multimodal support, structured output, and tool calling, empowering scalable AI workflows for thousands of developers. • Ranked #1 contributor at langchain-google for three consecutive months, merging 19 PRs and adding 5,000+ lines of code, directly improving core framework functionality. • Collaborated with maintainers and the community to develop comprehensive documentation and workflows, increasing accessibility and adoption of advanced AI tools within the open-source ecosystem.
• Developed and deployed a real-time fire and smoke detection system using YOLOv8n, improving emergency response times by 40% and reducing false positives by 30% in industrial settings. • Integrated image augmentation and backend systems for enhanced detection in challenging factory environments, providing a scalable solution for real-time monitoring and alert management.
• Developed an LLM-powered chatbot with a comprehensive document processing pipeline using Azure OpenAI, LangChain, Ollama, and Nomic embeddings, enabling real-time PDF analysis and intelligent query answering, reducing academic query response times by 60%, enhancing user interaction by 30%, and improving operational workflows at Habib University.
• Designed and validated machine learning models to forecast agricultural commodity prices with 95% accuracy, enabling data-driven decision-making for stakeholders. • Automated data pipelines for scraping and preprocessing, reducing manual effort by 99% and enabling real-time updates within 5 minutes, boosting operational efficiency by 30%.