I am Rohan, an experienced Ed-Tech Product Manager and a recently graduated Learning Engineer.

Masters from Carnegie Mellon University's METALS
(Masters of Educational Technology and Applied Learning Science) program.

5+ years of product management experience in two ed-tech start-ups, Funtoot and Educational Initiatives.

Interested in personalized learning, intelligent tutoring systems, educational data mining, skill acquisition, expertise and lean entrepreneurship.

Exactly 2 years after undergrad, I crashed my startup, DewInk, but realized I loved being in the classrooms. Read more about me and my story here.

Goals for Masters Program

1) Build a cognitive model in ACT-R, a cognitive architecture.

ACT-R is a cognitive architecture: a theory about how human cognition works. The architecture strives to understand how people organize knowledge and produce intelligent behavior. In the Cognitive Modeling class by Prof. John Anderson, I wrote ACT-R models in LISP for a multicolumn addition task and the speed-up associated with recall while repeatedly doing alphabetic arithmetic task e.g., A + 2 = C.

2) Build a production rules tutor.

Cognitive skills are realized by production rules. Each production rule is a modular piece of knowledge and complex cognitive processes are achieved by stringing together a sequence of such rules. This signifies that production rules are the units of the skill or simply stated, a skill is learnt in pieces. In the Personalized Online Learning class by Prof. Ken Koedinger, I built multiple production rule tutors, starting off with a simple shortcut strategy tutor for finding squares of numbers ending in 5, followed by a more nuanced proportional reasoning tutor for grade-6 & grade-7 math and culminating with a Menedelian Genetics tutor.

3) Conduct a Cognitive Task Analysis (CTA) of a math domain that has troubled students.

Cognitive Task Analysis (CTA) is the use of structured knowledge elicitation techniques to discover underlying cognitive processes, rather than observable behaviors associated with a task performance. Symbolization (a skill wherein an algebraic word problem is given and a symbolic expression is desired) has always been an achilles heel for grade-9 and grade-10 students. My Symbolization CTA was a think-aloud conducted with a novice and an expert and outlines a cognitive model driven by building of quantitative sub-systems.

4) Build an educational game backed by evidence based e-learning principles.

As part of Design of Educational Games class, I coded Haunted Path - a game to teach decimals through a number line. The game development included a CTA, paper prototyping, game engines exploration, coding in GameMaker Studio and a final evaluation. The game trailer highlights the mechanics and the e-learning principles involved such as immediate feedback, spaced out practice etc.

5) Envision a product using human-centered design approach.

Human-centered design is a process that places human needs, desires, and limitations above other factors. As part of Interaction Design Overview, the 5-week assignment saw us accomplish the following steps; user interviews, affinity diagrams, storyboards, personas, customer journey maps, paper prototypes, user testing and led us to PhotoJog - an app to restore lost connections between patients and their family caregivers by driving joyous conversations around old nostalgic photos. The detailed report outlining the process and all the intermediate steps is here.

6) Understand how intelligent tutoring systems (ITS) function with respect to their inner and outer loops.

Intelligent tutoring systems (ITS) have two loops, an inner loop for step-based feedback and an outer loop for problem selection. The inner loop executes once for each step taken by the student in the solution of a task whereas the outer loop executes once for each task, where a task usually consists of solving a complex, multi-step problem. My implemented Symbolization Tutor is an ITS that,
Extends the inner loop functionality through the use of follow-up questions based on a strategy. Modifies the outer loop problem selection algorithm to give isolated practice by selecting problems with the fewest unmastered skills.

My foray into educational technology was accidental. Exactly 2 years after undergrad, I crashed my startup, DewInk, a visual storytelling community for children and the lone silver lining was a realization that I loved being in the classrooms.

This led me to Educational Initiatives (EI), an exciting ed-tech startup headed by a visionary founder, Sridhar Rajagopalan. I stayed at EI for close to two years and realized two things-classroom learning is largely rote and technology alone cannot solve any nation’s learning woes.

During the first few months itself, EI nudged me inside classrooms to conduct some teaching sessions. I used these sessions to uncover the underlying reason behind common student bugs or misconceptions e.g. why do students associate angle measure with the length of the arms? why are students unable to measure length of a pencil using a scale, when the pencil tip is at 1 and not at 0? why do students place ¾, a fraction less than 1, between the numbers 3 and 4 on a number line? Most of the derived insights were later fed into Mindspark, EI’s adaptive math learning product.

By the time I left EI, I realized my real passion lay in building large scale personalized learning systems. I applied and got selected for LearnLab’s summer school-an intensive 1-week course focused on creating technology-enhanced learning experiences and building intelligent tutoring systems at Carnegie Mellon University (CMU), Pittsburgh. This was truly an enlightening experience, there were lectures and workshops on ACT-R, production rules tutors, bayesian knowledge tracing, knowledge component analysis using Datashop etc. My final project was with Dr. Ryan Baker on ‘class level differences in student bugs’.

Post LearnLab’s summer school, I joined a very young startup - Funtoot, as a Product Manager. Funtoot was attempting to create India’s most cutting edge adaptive learning system for Grade 2-10 Math. As of today, Funtoot is a multistep, feedback based learning system, currently being used by roughly 120,000 students across 200+ schools.

While EI drove me more towards science of learning, Funtoot made me a robust Product Manager-the kind that can effortlessly manage user needs by not necessarily being a feature junkie. In the short span of 2 years, we shipped four different products-a better version of Funtoot, a content authoring tool, a parent app and a teacher interface with reporting, revision and test functionalities.

I joined CMU’s METALS program last August to become a Learning Engineer architecting empowering learning solutions.