Timely and consistent feedback is critical to student success, yet grading written assignments remains a labor-intensive process for instructors and teaching assistants. This project introduces an AI-powered grading system that integrates OpenAI’s GPT-4o-mini API through a function-calling orchestration architecture. The system consists of a central AI agent that coordinates specialized tool modules—such as the Submission Fetcher, Rubric Retriever, Criteria Interpreter, and Feedback Generator—each executing distinct but interdependent stages in the grading pipeline. These tools collaboratively perform PDF document retrieval from storage, rubric-aware interpretation with contextual enhancement, and structured feedback generation through carefully engineered LLM prompting with temperature-controlled consistency. All AI-generated grades undergo review before final submission to the Canvas Learning Management System (LMS), with version tracking and audit trails maintaining full transparency of the automated decision-making process with instructor-friendly User Interface. This approach improves feedback speed and grading consistency, aiming to save significant time for instructors and TAs while enhancing student engagement and academic outcomes through faster, more detailed feedback delivery.
Watch the team present this project at 01:41:30 in the session recording here.
Keywords: large language models (LLMs), auto-grading assistant; multi-agent systems; auto-grading assistant; rubric-aware scoring; Canvas LMS integration; OpenAI API; educational AI; instructional technology; human-in-the-loop learning analytics
Faculty Advisor
Dr. Ali is a seasoned leader in data and analytics, ascending from an engineer role to senior leadership. He has successfully led, designed, and built some of the largest global data solutions across diverse industries such as telecom, financial services, insurance, CPG, Retail, and manufacturing. His expertise spans leadership, product/program management, data engineering, data architecture, data security, AI/ML ethics, and privacy.
His successful global career includes roles across Asia, the UK, EU, MEA, APAC, and the USA. He gained extensive experience at major organizations such as NCR, Teradata, Barclays Bank, Capgemini, and Circana (IRi). His client portfolio is extensive, featuring top global companies like Telenor, Vodafone, Morgan Stanley, Citigroup, The Hartford, CNA, Altria Group, Walmart, and Apple. After working at large corporations, he transitioned to the start-up ecosystem and worked at two AI unicorns i.e. Dataiku and Sigma Computing and later co-founded two AI start-ups i.e. Realix.ai and Digital Emissions.
He now focuses on education and paying forward. He teaches in graduate data science programs and is a regular speaker on data and AI panels and a frequent guest lecturer. He remains active in the entrepreneurial space, advising early-stage start-ups on product innovation, leadership, culture, customer success, and growth strategies. He is a perpetual learner and holds multiple advanced degrees and continues to learn every day.