Harsh Satija

me

Background

Research

I'm mainly interested in Artificial Intelligence and Machine Learning, particularly Reinforcement Learning and Deep Learning. For now, my research project is rooted around using RL in the domain of NLP.

Curriculum Vitae

A copy (updated on Nov 2015) is available here. Contact me for a more recent version.

Publications

Experience

  • Data Scientist, Sokrati (Dec 2014 - Jun 2015) : I had a short gig at India's leading Digital Advertising firm where I built predictive models for forecasting, bidding and budget allocation for a large number of clients across domains like e-commerce and banking. The algorithms manage advertising portfolio of millions of ads in completely automated fashion. I also developed recommender systems for user acquisition and retention in retail space.
  • Software Developer Engineer, Amazon (July 2013 - Dec 2014) : I built a tax identity management service which enables Amazon to be US Tax laws compliant entity via collecting legal data of the merchants, validating their identities and then serving as authoritative source of tax info to other services and businesses internal to Amazon.

Teaching

I have been a Teaching Assistant for the following courses :
  • Applied Machine Learning, COMP-551 (Winter 2016) : Course covered an overview of ML algorithms with focus on applications, targeted at grad students. Duties included helping students, running tutorials, designing and grading assignments.
  • Artificial Intellligence, COMP-424 (Winter 2016) : Course covered basic AI algorithms and methodologies, targeted at CS majors. Duties included helping students, running tutorials, grading assignments.
  • Intro to Software Systems, COMP-202 (Fall 2015) : Course covered Bash, basic UNIX tools, C, and basic UNIX programming. Duties included helping students, marking assignments and exams, and designing an online judge for automatic evaluation of programming assignments.
  • Algorithms (Fall 2012) : Course covered algorithms relating to various data structures, sorting,algorithm analysis, memory organization, and dynamic memory. Duties included helping students, testing students in person, exam invigilating, and running review sessions along with Intructor Prof. Kannan Srinathan
  • Science II (Winter 2012) : Course covered introduction to quantum mechanics, theory of relativity and exposure to quantum computing. Duties included designing quizzes and grading assignments along with instructor Prof. Harjinder Singh

Projects

Most of my projects are unfortunately private and I'm in no liberty to release them. A few of my projects can be found at my github (but they are way too old - pre 2012). Though, a few of my recent course project reports can be found here:
  • Variance Reduction Techniques for Sequential Structured Variational Inference [Project Report]
  • Scaling Deep Q-Networks [Project Report]
  • Optimality Bounds in Reinforcement Learning from KWIK Perspective [Project Report]
  • Using Disambiguated Word-embeddings for Exploiting Similarities among Languages for Machine Translation [Project Report]

Other Interests

I enjoy dancing (both street - Locking/Popping and ballroom - Salsa/Tango/Bachata/Cha-cha), learning new languages and cooking (better). I recently discovered K-pop, both dance and music, and currently hooked to it.
I I enjoy vast genres of music and I listen to anything from Beatles to Tool. I think these days my taste in music can be generally biased towards Melodic Death Metal and Progressive-Metal (along with small doses of Classic rock, Alternative, Djent, Progressive rock and Folk-metal ). You can find me on Spotify to check out to which bands I'm currently hooked to.

Contact

  • Email (the best way to reach me) : harsh.satija with the suffix (@mail.mcgill.ca)
  • Office Address : McConell 106, School of Computer Science, McGill University, 3480 University Street, Montreal, Quebec, Canada, H3A 0E9
  • Postal Information: McConnell Engineering Bldg. Room 318, 3480 University, Montreal, Qc, Canada, H3A 0E9
Template adapted from Eric Schulte