Hello! I'm Aadil Hayat. I am currently a M.Tech. student at IIT Kanpur. I am very passionate about technology and programming. I am interested in pursuing higher studies in Machine Learning.
M.Tech. (Computer Science)
Indian Institute of Technology Kanpur
Indian Institute of Technology Kanpur
Delhi Public School Kalyanpur, Kanpur
High School (ICSE)
Margaret Leask Memorial English College, Jhansi
InfoRL: Interpretable Reinforcement Learning
Developed and implemented InfoPPO algorithm which uses information maximization to learn interpretable latent codes. Ran the experiments on OpenAI Mujoco environments to learn latent codes controlling varied configuration over trajectories. Trained policy can control speed, configuration, etc for Mujoco control tasks like Walker, Humanoid, etc. Reduced the problem of over-fitting in multi-agent self-competition learning using this algorithm.
Multi-Task Learning using Adversarial Imitation Learning
Implemented Generative Adversarial Networks for Imitation Learning on complex RL environments. Implemented an algorithm to learn a single policy to perform multiple but correlated tasks in Mujoco environments. Implemented successful imitation of 2 different driving (pass and turn) tasks in TORCS environment. Paper accepted at NIPS’18 Workshop on Imitation Learning and its Challenges in Robotics.
Probabilistic Machine Learning
Implemented 2 novel approaches to generate adversarial samples to attack Deep Learning based classifiers Used Generative Adversarial training inspired by GANs to learn a generative model to attack Neural Networks
Transfer Learning using RL agent
Implemented and trained Asynchronous Advantage Actor-Critic Method based agent for DOOM environment. Used the trained Convolutional layers of the agent to perform Computer Vision tasks like object detection,etc
Implemented continuous control using actor-critic based deep reinforcement algorithm in Keras. Trained a helicopter agent to learn different tasks from Reinforcement Learning Challenge 2014.
Deep Population Genetics
Research Internship NYU
Implemented pre-processing of genetic data from FSTAT format to allele-frequency matrix. Implemented dimensionality reduction by Auto-encoders using Tensorflow and Keras on HPC cluster. Implemented optimal number of clusters prediction using t-SNE, Silhouette coefficient and k-means clustering. Algorithm is used for deciphering organization of populations in space and time using genetic data.
Implemented foreground-background separation using temporal median of pixels in video. Implemented video synopsis using Markov Random Fields. Solved Markov Random Fields using iterative graph cuts and loopy belief propagation.
IITK Surveillance Video:
Object Detection and Classification
Machine Learning Techniques
Implemented foreground-background separation of surveillance video using Gaussian blur and threshold dilation. Implemented object detection by finding contours in background separated video. Implemented object classification by bag-of-visual-words model using SIFT descriptors.
Neural Networks Based Chatbot
Natural Language Processing
Implemented Google's Seq2Seq model for conversations using TensorFlow library for deep learning on Nvidia GTX 760 GPU.
Principles of Programming Languages
Developed Interpreter for Kernel Language of Mozart-Oz in Oz. Implemented parsing of AST form of code and closures of functions
Web Infrastructure and Crawling
Installed and configured CoreOS and Dockers on a bare metal server remotely. Developed cloud-config for CoreOS and systemd unit service files for various dockers. Developed custom Dockerfiles for Postfix, MySQL and Hadoop cluster. Developed DOM parsers for Organizations' and Individuals' pages from professional networking websites. Implemented multi-threaded python program to crawl specific websites.
Inter IIT Tech Meet'15
Developed a system to make the popular content available on phones having sporadic data connection. Designed backend server, implemented automatic modification detection and incremental updates to client. Implemented intelligent storage management, file downloading and file rendering system client side.
Developed 3d version of Youtube using ThreeJS and added gesture support using LeapJS and LEAP controller. Implemented gesture based scrolling, play/pause, mute/unmute and quality change. Designed intuitive and attractive 3d user interface and implemented 3d animations.
It's Placement Time
Implemented OpenGL API in C++ to develop a 3d game, a comical picturisation of placements. Implemented gravity simulation, multi-player gaming, heads-up display, texture mapping and saving screenshots. Video presentation of the project can seen here.
Developed a web application with the attempt to incorporate all the online services available for IITK students. Scrapped students, course, and exam schedule databases of IITK having 3000+ results. Developed intuitive UI for student search and exam schedule.
Developed web application for online programming competitions. Designed Administrative tools for intuitively managing and conducting competitions. Implemented background program compilation and execution for C and C++.
Fundamentals of Computing
Developed Windows Store App for events management at IITK. Developed JSON API for event synchronisation between database and local clients. Designed Intuitive UI for event creation and notification
Other Technical Projects
- Can Crusher
- Pin Code Explorer
- IITK Timetable App
- Facebook Timeline Search
- Electronic Notification Board
Frameworks & Libraries
Tools & Packages
Areas of Expertise & Interests
- Placement offer from FlowTraders, Amsterdam for Trading Support Engineer
- Secured All India Rank 1729 among nearly 1,500,000 students in JEE Advanced 2013
- Secured 3rd position in Open Soft (Software Development) competition at Inter IIT Tech Meet 2015
- Awarded 2nd Project Appreciation Award in TA-202 Manufacturing Lab 2014-15
- Secured 3rd position in Hackathon during Enterpreneurship Summit 2014 organised by E-Cell IIT Kanpur
- Secured All India Rank 42 in National Talent Search Examination 2012 conducted by Unified Council
- Watch Movies
- Playing Badminton