May 2020 – Aug 2020 · 3 mos · Palo Alto, California, United States · Remote

• Developed pipeline using OracleDB and Perforce API, Python, and Pandas to extract test results data and analyzed continuously.
• Successfully carried out an analysis on test results data to correlate files and workloads failure, which helps reduce triage time drastically.
• Synthesized formulae to give failure score to workloads given the list of files edited to identify workloads most likely to fail.
• Optimized data mining operations using Pandas and distributed operations over multiple cores using Python Multiprocessing.

Jan 2020 – Mar 2020 · 2 mos · Raleigh-Durham-Chapel Hill Area

• Created a data pipeline in Python that extracts 2 million individuals’ hospital records from PostGresDB and outputs Boolean attributes.
• Leveraged Pandas to manipulate huge data and distributed operations over multiple cores, decreasing the processing time by 30%.

Jan 2019 – Jun 2019 · 5 mos · Pune, Maharashtra, India · On-site

• Devised an algorithm in Python to differentiate OS processes from the application process, reducing user effort to get safe processes.
• Introduced novel clustering algorithm to group VMs that run similar processes and make similar connections with 95% accuracy.
• Collaborated with DevOps team and automate the job of getting information on open source libraries leveraging Jenkins pipeline.
• Developed the shell script which setups the virtual machine and builds the AppDefense project when executed by Jenkins job.
• Analyzed the patterns in DLLs when the process starts to identify the DLL injection attack by finding uncommon DLLs.

Jun 2018 – Dec 2018 · 6 mos · Ahmedabad Area, India · Remote

• Research and studying financial trading algorithms and their practical use against use cases. Made contributions, while working with our team members in the following areas:
• 1. Research, identification, and selection of ‘best fit’ algorithm(s); LSTM, SVM, CNN, etc.
• 2. Designing and validating the machine learning model(s);
• 3. Testing and training models, identifying areas for improvement; and
• 4. Implementing change.
• Designed states from the time series on Market index data that significantly reduced noise when fed to the DL model. Designed trading strategy for a rule-based system that handles uncommon data when used with the DL model.
• Designed stock deep learning model pipeline that extracts data and displays technical indicators in real-time.
• Introduced a data preprocessing technique that eliminates 65% of false trade signals by converting intraday data into character sequence.
• Designed and implemented sequence to sequence CNN-LSTM with attention mechanism to predict future sequence for a trade signal.
• Worked on building a rule-based model with a machine learning model to increase the accuracy by manually coding strategies.

May 2018 – Jun 2018 · 1 mo · Ahmedabad, Gujarat, India

• Involved in research work.
• The main topic of research was to train multiple faces and recognize in real-time with the best possible accuracy.
• Building a surveillance system for generic use.
• Build one of the company’s client projects. The objective was to build backend web service for unlocking the phone by face.
• Challenges such as reducing false positives and giving quick results were achieved.
• REST API was created for the application with a response in JSON.
• Developed algorithm using OpenCV HOG and Dlib face encodings which concurrently train new faces while predicting known faces with false-positive of only 0.6%. Developed API using Flask which gives results in 2.8 seconds for 60 fps video and 20 faces in the database.

Aug 2017 – Aug 2017 · 0 mo · Gujarat, India

• Learn web designing using bootstrap frame work and deployment of project