Deep Patel

-Fun project at office hackathon at LiveIntent.
-Extracted keywords from HTML templates using NLP.
-Used Word2Vec model to find advertisement most similar to the context in a news template.
-Used Google TensorFlow to get keywords from images.
-Auction based model where the advertisement most suitable to the context wins based on basic scoring.

-Collected a data set of 10000 Wikipedia HTML documents.
-Using Beautiful soup (Python library) extracted the text from the HTML documents.
-Using gensim's word2Vec with Google's News Vector Model, classified Wikipedia documents into 25 categories.

Implemented algorithms for DNA sequence alignment - Needleman Wunsch, Smith Waterman, Hirschberg's algorithm.
With varying input data sizes, experimented and plotted the graphs.
Finally, most importantly, Analysed the algorithms to understand their complexities- both space and time.

Built a webpage for predicting long and short term stock prices using SVM, artificial neural network,
Bayesian algorithms giving recommendations like Buy, Hold and Sell for each stock.
Fetched stock data from Yahoo API and stored in MySQL database. Developed dynamic stock graphs and indicators graphs using D3.js.
Implemented Features like Live stock Ticker, adding stock to favorite and tracking, comparing stocks.
Added news feed to the webpage using RSS newsfeed.
Scripting with… Show more

Built a webpage for predicting long and short term stock prices using SVM, artificial neural network,
Bayesian algorithms giving recommendations like Buy, Hold and Sell for each stock.
Fetched stock data from Yahoo API and stored in MySQL database. Developed dynamic stock graphs and indicators graphs using D3.js.
Implemented Features like Live stock Ticker, adding stock to favorite and tracking, comparing stocks.
Added news feed to the webpage using RSS newsfeed.
Scripting with Python with Django web-framework. Developed front end using HTML, CSS, JavaScipt and Jquery. Show less

Built a system that consists of a sensor node to sense the room temperature, display on LCD, wirelessly transmit temperature using CC2500 wireless modules to base station computer which logs all the values.

Implemented a digital level indicator using the PSoC boards. The train of LEDs was used as the level indicator as a substitute for the bubble in the spirit level indicator. I interfaced the accelerometer on the PSoC board to check for the surface level.

Built a air mouse that needs no base to operate using mbed FRDM KL25Z embedded systems board. The cursor works on the basis of inclination of mouse in air as the on baord accelerometer transmits values via serial-USB interface. The right and left click were implemented using capacitive touch sensor on the board.

• Applied Deep Neural Network for recognition and classification of accelerometer mobile data into different activities
like walking, running, climbing etc. using R on Spark.
• Machine learning algorithms like random forests and gradient boost machine were applied for increasing the accuracy.
• Spark H2O framework was used to parallelize the algorithms over IBM Bluemix cloud platform for faster computation.

Implemented K-means Algorithm using Python for analysis of Lending Club data to cluster the users based on their FICO scores to analyze the risk associated with each user.

Using C++, built a stock portfolio bank system that allows user to buy and
sell shares, view transaction history, view graphs of share price variation and user’s stock portfolio.
Built a bank system to withdraw and deposit money.
The system logs and prints user’s Stock Portfolio and Transaction history with date and
time.

Using twitter API, fetched half million live and historical tweets of users all over USA, stored in MongoDB NoSQL database.
Using Data analytics, statistics were developed to compare the intensity of health related
activities in each state.
Created Dashboard to display all Data analysis and results.
Technologies used: Python, Django, HTML5, JavaScript, CSS

Built a 2-motor robot that moves with the gestures of our hand, while feeding live video to the control center.
We used Mbed FRDM KL25Z on Embedded systems platform and wireless ZigBee modules for wireless transmission.
We also controlled the speed of the bot by using PWM.
C language was used to code the Mbed boards.

As an intern at Iboon Technologies, worked as a summer intern on a Software project to build a "Appoinment Scheduler".
It was built to help doctors schedule appointments for patients. Admins can add clinics and doctors.
The patient gets a notification one day before, and on the day of appointment.
Technologies used: Java, HTML5, JavaScript, CSS, SQL