Shubham Patel
I am a 3 years experienced developer, currently pursuing my Master's in Computer Science from Stony Brook University. Passionate about obtaining a position that helps me evolve both in technical and professional front while achieving organizational goals in parallel.
Experience
Application Engineer
Worked in DevOps PETAP team under OAL (Oracle Application Labs) on Expenses and Travel module, under both Oracle E-Biz and Fusion (Oracle Public Cloud). Implemented and released to production, custom enhancement, to make expense and cash advances submission, approval, audit and payment process efficient for ~138000 Oracle employees.
Teaching Assistant
CSE 114 - Introduction to Java, assisted professor in data logging in reporting system, organized and conducted review session for students and helped in grading exams and homework assignments.
Education
Stony Brook University
GPA: 3.46
Motilal Nehru National Institute of Technology
CPI: 7.96
Skills
- Python
- Java
- C++
- C
- PL/SQL
- Oracle DB
- OBIEE
- Java Script
- HTML
- CSS
- APEX
- OJet
- Angular JS
- Node JS
- React
- Oracle Forms
- Numpy
- Pandas
- Pytorch
- Keras
Projects
IoT Security and Verification - Building Vendor Independent Models for IoT Applications
Ongoing project, involves creating VI graph bases model (OpenHab, SmartThings, MUD, IFTTT supported so far). Model detects conflicts/violations, verify it’s correctness and automatically resolved them.
Visual Odometry with Deep Recurrent Convolutional Neural Networks
A typical VO pipeline consists of camera calibration, feature detection and extraction, feature matching, outlier removal (RANSAC), motion estimation and local optimisation. All these steps require significant engineering effort and also need fine tuning as per implementation condition. Bypassing this tedious engineering effot, we implemented an end-to-end framework for monocular visual odometry (VO) problem by using the deep RCNNs, using KITTI monocular VO dataset.
Image Classification using bag of visual words
Improving baseline tiny image model using bag of SIFT features
Involves KMeans clustering for features, KNN and SVMs model for classification.
Action Recognition
Action recognition and classification among 16 classes by analyzing video frames.Pretrained VDD16 was used for feature extraction and action was recognized/classified using both SVM and LSTM RNN network, accuracy 95.77% and 95.59% resp.
Personality Driven Chatbot
A chatbot trained on adapting conversation style of a certain speaker.
Trained on personal chat and character Ross from FRIENDS TV sitcom. GENSIM and GloVe were used for word embedding generation. Model architecture involved BLSTM (Keras), LSTM with attention model (Pytorch).
Match Constraint Sequence Alignment
Algorithm to accelerate the performance of the genome sequence alignment,by using a combination of exact and selective alignment, with a cache
Results prove algorithm is faster than existing KSW2 library
Human Emotion Detection via Facial Expression Recognition
Detects emotion : Happiness, Fear, Sad, Anger, Surprise from face expressions
Cohn-Kanade DB was used for training SVM model, 83% accuracy
NETMIN & WEBMIN
Created a user friendly terminal interface (NETMIN)/web interface (WEBMIN) that allows user to perform general tasks like user management , file management , apache server handling, dns server management , port management, both remotely and smoothly
Awards & Certifications
- OAL FY17Q2 Award of Excellence, 2017
- Achievement Certificate for NETMIN Project, Nettech India, 2013
- Winner of SOFTATHON and OPERAOMNIA, Cyberquest MNNIT, 2013
- Global Rank of 214, International Maths Olympiad, 2009
Interests
Apart from my technical facet, I am an ultrarunner, a cyclist and an aspiring triathlete. I realized that endurance sports fuel my chi, helping me focus on my immediate goals and being more productive. My enthusiasm towards fitness drove me to be a fitness blogger and mentor people under ‘Couch to 5K’ programme. This way I can inspire people to become better versions of themselves and realize their true potential.
- Hyderabad Ultra Marathon (50K) 2017 [07:08:11]
- Hyderabad Olympic Duathlon 2017 [03:55:20]
- Standard Chartered Mumbai Half Marathon 2017 [02:08:43]
- Hyderabad Club Run, Haf Marathon 2017 [01:58:20]
- JIO Amravati Half Marathon 2017 [02:07:02]
- Airtel Hyderabad, Half Marathon 2017 [02:22:21]
- Shriram Properties Bengaluru Half Marathon 2017 [02:06:29]
- Hyderabad 10K Run 2016 [51:24]
- Self supported Cycling Brevet from Hyderabad to Hampi 2018, 380 kms, 39 hrs [Active time - 22 hrs]