Islāmābād, Pakistan
MS in Computer Science from Iowa State University (4.0 GPA), with a focus on artificial intelligence, advanced algorithms, and core computer science topics. Engaged in AI since 2018, leading to research roles, projects, a publication, and 3 years of full-time work as an AI Engineer. * The views expressed on this platform are my own and may not necessarily reflect the views of my company or any of its officials.
• Leading development of a natural-language agent with tool-calling capabilities for automated catering order processing • Overseeing a voice-agent for inbound sales/demo-booking calls, integrating conversational UIs and booking systems
• Improved contextual relevance in Ephor.AI's (https://ephor.ai/) LLM responses by implementing a RAG system with Pinecone and ReadAI meeting data • Engineered a companion macOS app for Ephor.AI using PyQt and py2app, compatible with Intel and Apple hardware • Developed an AI-powered productivity monitoring feature leveraging AWS Bedrock to aggregate user activity data and deliver personalized performance insights • Reduced finance data collection workload by 80% by developing a Playwright-based CDR extraction solution on AWS
• Developed a distributed RAG system on AWS infrastructure, optimizing vector search for enhanced performance and scalability • Engineered a sensitive information detection system combining a fine-tuned T5 transformer with conventional methods • Implemented a RAG-based chatbot for company policy queries with 90% accuracy by optimizing key RAG components • Engineered prompts to generate reports evaluating startup success potential based on multiple metrics
• Trained and optimized a YoloV7 model for real-time object detection on an Nvidia Jetson Nano • Achieved a 32% reduction in model size and 7x increase in inference speed using TensorRT quantization • Designed and implemented a novel crop row detection and vehicle heading error calculation algorithm • Conducted exhaustive literature reviews and presented insights on pivotal ML/AI tools and practices
• Improved object detection model performance by 32% by preprocessing data and tuning its distribution • Detected 4 safety violations with 78% accuracy by writing custom logic to process model detections for CCTV footages • Clustered medical procedure details (10,000+ items) using features from NLP models and vector similarity measures • Decreased video processing time by 240% by incorporating multi-processing into a video highlight generation program • Analyzed Covid-19 data to obtain valuable insights and created informative visualizations for technical reports • Productionized backend for web application to create custom tabular datasets supporting 8 distributions • Developed a high-performance image stitching algorithm to generate composite images from smaller constituent parts • Created presentations and tutorials introducing data science and machine learning to 100+ aspiring data professionals