Career Profile
Software Engineer at Bloomberg LP, London. Working with Yield Curves and Fixed Income Pricing. I help build and maintain Software Infrastructure for generating and publishing relevant data. I write code in Python and C++. My research interests lie in Machine Learning, Artificial Intelligence, Reinforcement Learning, and Quantitative Finance
I have worked as an Intern at Siemens Corporate Research, Bangalore India during Winter ’16 and Summer ’17 in the field of Reinforcement Learning and Deep Learning. I have also pursued research in Deep Learning during my Bachelors study at IIT Mandi.
Experiences
Working with BVAL (Bloomberg Valuation) which deals with Fixed Income pricing using quantitative models. Primary languages are Python, C++ and JavaScript
TA for AI course taught by Prof. Deepak Khemani.
Intelligent Traffic Optimization using RL. Developed and simulated the system on a small segment of Electronic City, Bangalore
CS201 - Computer Organisation taught by Dr. Anil Sao.
Major duties were evaluation of lab assignments and clearing doubts
Built a Reinforcement Learning agent that was used to automate few rooms of a building simulator. It was designed to optimise energy usage.
Honours & Awards
Awarded to the student with the best academic record among the graduands.
Judges award for the best Major Technical Project titled "Proactive anomaly detection in resource usage metrics using LSTM and other methods".
Developed a system to analyse trend of a particular stock and scrape news articles(Previous two days) that could explain it. Technologies used were Python, Flask, NLP(Sentiment Analysis and Document Summarisation), Beautiful Soup. Project can be found at this Github repo
Insreasing seatbelt awareness among Himachali drivers by raising awareness, conducting surveys and developing an IOT device to track real time compliance.
Projects
A Novel LSTM based model which uses "Association Learning" was used to predict resource usage in cloud machines. Model was validated across multiple datasets and found to perform better than other state-of-the-art models.
Technologies used: Keras, Tensorflow, Python, Docker
Published at: 2018 International Joint Conference on Neural Networks (IJCNN)
A two stage novel model was built where in 1st stage, hybrid of long short term memory (LSTM) and bidirectional long short term memory (BLSTM) is used to predict future resource usage. In 2nd stage, LSTM classification model was build to classify expected state as normal or abnormal.
Technologies used: Keras, Tensorflow, Python, Docker
Published at: 2017 14th IEEE India Council International Conference (INDICON)
A modified kernel that can allow user to encrypt files on-the-fly, so overall user experience remains seamless whereas actually the file remains encrypted on the disk.
Each data block is encrypted and stored on write() and decrypted on read(). Kernel read() and write() functions were modified for this purpose. Encryptor was tested on ciphers where size of encrypted data may differ from that of plaintext.
Technologies used: Linux kernel, Linux Virtual File System
Clustered MariaDb was used as the database managemnet system. Database was tested on exact search queries, range queries, index queries, etc. Indexing was done on the database to enhance query performance.
Technologies used: MariaDb, Galera cluster
This was a social as well as technical project. Several surveys were conducted to identify root cause of lack of seat belt awareness. Built a prototype which can track whether passengers in car are wearing seat belts. The seat belt usage data is futher sent over cloud. The data on cloud can be accessed by governing agencies to track seat belt usage.
Technologies used: IR sensors, GSM modeule, accelerometer, MySQL
IOT device for monitoring usage of LPG gas cylinders along with leakage alerts through Email and SMS
A web portal for users to view and book institute vehicles (cabs, buses, etc.)
Technologies used: MySQL, PHP, LAMP, JavaScript, HTML, CSS
The portal provides facility for online secure money transactions, creating accounts and checking balance.
Technologies used: MySQL, PHP, Python, JavaScript, HTML, CSS
Arduino and Piezo sensor based lock. Piezo sensor was used to detect the vibrations. Used Arduino programming for input processing.