Shashank Shekhar

Shashank Shekhar

AI Resident at Facebook AI Research

Portfolio

Label Key
Language Library API Project type Subject Area

Nand2Tetris

GitHub
HDL Computer Systems

Implemented a a full fledged computer called HACK from scratch: logic gates, MUX, DeMUX, Adders, ALU, 16-bit registers, 16-bit PC, 8/64/512/4K/16K RAM, Memory (RAM+ROM+peripheral), CPU (1 bit op-code and 16-bit intruction set). Also wrote an assembler to translate HACK machine code to 16-bit instructions executable on the HACK computer.

Flappy Crewmate

GitHub
Python Pygame Video game

Built a simple clone of the Flappy bird game using character sprites, backgrounds and sound effects from the popular Among Us game.

Probabilistic Programming with Pyro

GitHub
Python Pyro Research Bayesian Inference

Wrote an introductory set of jupyter notebook tutorials on probabilistic programming, inference, MCMC, variational infernece and variational autoencoders using Pyro

RBDoom

GitHub
Python Pytorch VizDoom Reinforcement Learning

Implemented a reinforcement learning based agent to play the first person shooter game DOOM using only image-pixel inputs. Used the RAINBOW algorithm based on deep q-learning and the VizDoom emulator for the game.

Meta-Learning with Implicit Gradients

GitHub
Report
Python Pytorch Research Meta-Learning

Implemented the NeurIPS 2019 paper using the Torchmeta and Higher librarries for meta-learning and higher-order gradients and replicated the results obtained by authors for few-shot learning on Ominglot dataset

Cross-Modal GAN

GitHub
Python Pytorch GAN

Implemented the cross-modal GAN paper from "Cross-Modality Person Re-Id with Generative Adversarial Training" at IJCAI 2018.