About

Hello! I’m Krishna Agaram, a final-year CS undergrad at IIT Bombay. I enjoy reinforcement learning, robots, deep learning theory, pure math and dev projects in my areas of interest. I’ve worked and am working on reinforcement learning, property testing and quantum information. Oh, and of course, I’ve done a fair amount of dev throughout my undergrad. Pure math gets me pretty excited. Robotics Dev as well: it’s kinda the intersection of lots of stuff I like.

If I’m not working some math, I’m likely reading Asimov or Christie, playing the piano, or am outdoors on a run or bike, climb or hike. I’m always looking to learn new things and meet new people, so feel free to connect if any of the above topics interest you too!

My (old) CV, Github and LinkedIn.

I sometimes make notes, which you can find here. I am a big supporter of inquiry-based learning; if you’re interested, I encourage checking out JIBLM. I have authored one such guide to Linear Algebra and a more elementary, story-book guide to Counting (draft). Finally, I have a dead blog that needs revival.

Coursework

I was previously very theory-oriented, coming from a math background, and my coursework sort of reflects a bias in that direction. However, over the last year, I’ve realized that I get much more satisfaction when other people can benefit from what I do, and so I’m moving towards more applied stuff. Of course, math et al is still fun, and will remain a top hobby.

• undergraduate CS theory: data structures, algorithm analysis, logic, automata theory, cryptography, game theory and mechanism design
• graduate CS theory: quantum information, spectral graph theory, approximation algorithms, randomized algorithms, probabilistic proofs*, theoretical machine learning, statistical learning theory*, theory of deep learning*, formal methods in machine learning
• math: linear algebra, differential equations, calculus 1&2, analysis, abstract algebra, probability and statistics, numerical analysis, galois theory*, fourier analysis*, extremal graph theory and graph regularity (graduate)
• undergraduate systems: software systems, architecture, networks, operating systems, databases, compilers, advanced compilers*

(* – ongoing)

Misc

Some remarks (which certainly shouldn’t be on the main page, I apologize)

• One of my all-time favorite topics is Analytic Combinatorics, an area that marries combinatorics to algebra and complex analysis. It’s a beautiful book, built upon the composition of seemingly elementary operations applied to classes of combinatorial objects to build new ones, and I highly recommend it to anyone interested in generating functions. My dream is to one day have an opportunity to use it in CS theory research.
• Recently, after a course on theoretical machine learning, I have been giving some vague thought to a complexity-theoretic analysis of machine learning approaches: is it possible to have a complexity theory of deep learning, with problems falling into different (parameterized by size, surely) complexity classes corresponding to (say) different architectures? Essentially, if we think of machine learning as a new paradigm of computing, then this is simply its complexity theory. I’m not sure if this is a well-defined question, but it’s interesting to think about.