#342 Andrew Thangaraj: The $5,000 IIT Degree: Can India Fix Its Broken Education System?

What It Covers

Professor Andrew Thangaraj of IIT Madras details how the institute built a sub-$5,000 online BSc Data Science degree serving 40,000 active students, addressing India's broken higher education pipeline where only 27% of college-age youth enroll and quality degrees cost upward of $25,000.

Key Questions Answered

• •Skills-before-theory curriculum design: The IIT Madras BS Data Science program front-loads practical skills in its diploma phase, requiring students to build two functional apps and two ML models before advancing to theoretical coursework. This sequencing directly counters India's systemic gap where BTech computer science graduates often cannot run basic code on an actual machine.
• •Funnel-at-exit admissions model: Rather than filtering students at entry via hyper-competitive entrance exams, the program admits broadly — receiving 40,000–45,000 applications annually — then applies rigorous exit standards. Project completion acts as the primary filter, with 40,000 active students expected to produce roughly 1,000–1,500 four-year degree graduates and several thousand diploma graduates per year.
• •Instructor-layer scaling strategy: IIT faculty record lectures while a dedicated layer of full-time instructors — mostly PhD and master's graduates from top institutes who prefer teaching over industry — run eight hours of live online sessions weekly per course. This two-tier model separates content creation from student interaction, enabling enrollment scaling without overburdening research faculty.
• •Intermediate exit points enable low-income participation: A sub-$5,000 degree alone does not serve financially constrained students if it requires four uninterrupted years. The program builds in diploma and three-year degree exit points so students can enter the workforce mid-program, earn income through internships and part-time coding work, then continue toward the full degree without abandoning it entirely.
• •AI deployment in education hits a 10–15% error-rate wall: The program tested AI tools for automated code-review feedback across thousands of students and found error rates of 10–15% on specific coding tasks. At 10,000-plus users, that error volume becomes operationally damaging. Professor Thangaraj sets a 5% error-rate threshold as the minimum acceptable level before large-scale AI teaching-assistant deployment becomes viable.