//assets.cakeresume.com/assets/contentarea/contentarea-2d64eca8dae2a1ffb31e0fb1df887bbad072c7c19734d7d30fa14884bbcc299a.css

I believe I am a creative, and product-focused data science professional who enjoys developing end-to-end data-driven solutions in order to make the user or customer experience better. I enjoy working in a collaborative environment which generally brings out the best in others and myself, to address problems and brainstorm effective solutions, but I am also equally competent as an individual contributor. In the past, I have been praised for being a quick learner and having a problem solving attitude, with an ability to even work well during challenging deadlines.  

Professionally, I have 7 plus years of work experience, while working with real-world challenges, both in the industry and academic settings. In the data science area, I have been working with projects like personalised recommendations, time-series forecasting and classification, NLP use cases like sentiment analysis and text classification, computer vision projects like classifying images and detecting their quality, performing fraud detection, and developing interactive chatbots.


Major focus: Intelligent systems and Data Science

Embedded Machine Learning: 1-year long Project Group.

3 Research / Survey papers in Machine Learning area.

Projects:

1. Master Thesis - Deep learning for anomaly detection in autonomous systems
Realized state-of-the-art deep learning models to carry out anomaly detection on time-series data from the autonomous race car. Perform hyperparameter optimization, neural architecture search, unsupervised feature selection and create interactive visualisations

2. Predictive Maintenance in Autonomous Systems
Incorporation of ML algorithms for anomaly detection in an Industry 4.0 project - GMMs, Clustering, Decision Trees, ARIMA, etc

3. Embedded Machine Learning
Predictive maintenance for real-life industrial use-cases. Compressed the models and deployed it on Xilinx PYNQ board.

4. Sparsity in Neural Networks
Research on sparsity algorithms to prune and shrink large neural network models and run them on embedded platforms.

5. Face Recognition using Convolutional Neural Networks (CNNs)
Constructed and evaluated the state of the art models for face detection and recognition - LeNet, AlexNet, VGG, GoogLeNet, ResNet.

6. Chatbot using Natural Language Toolkit (NLTK)
Fabricated a naive chatbot trained on the Wikipedia corpus to respond to the user queries.

7. Face Generator using Generative Adversarial Networks (GANs)
Deep Convolutional GAN to generate realistic looking faces. Trained on the celebrities face dataset.

8. Neural Style Transfer
Compose images in the style of the artistic image using deep learning model.

9. Detection of Malicious URLs using Autoencoders
Realized a URL classification model design using Deep Autoencoder and filter suspicious URLs.

10. Credit Card Fraud Detection and Prevention using Deep Learning (DL)
Deep autoencoder trained on the Kaggle Finance dataset for credit card transactions use-case.

11. Sentiment Analysis of Trip Advisor Reviews: Natural Language Processing (NLP)
Used Bag of words model to analyze the sentiments in the feedback provided by the customers.

12. Facial Emotion Recognition: Inception Network and other popular CNNs
Face detection followed by emotion recognition using CNNs on the bounded face detected using OpenCV.

13. Object Recognition on Android: Tensorflow Lite Models
The model classifies the objects shown on the android device’s rear camera in real-time.

14. Simulating Self Driving Car: Udacity Nanodegree
Engaged with computer vision and deep learning to automotive problems, like detecting lane lines, predicting steering angles, etc.

15. Verilog HDL Compiler: Computer Science Project
Devised a compiler to achieve the lexical, syntactical, and semantic analysis of a Verilog HDL code.



University of Paderborn

Guru Gobind Singh Indraprastha University, India

Data Scientist

Nikhil Kumar Jha

TeamViewer GmbH

Senior Data Scientist

- Leading and coordinating development of different cross-team BI and Data Science projects.
- Develop, implement, and automating deployment of ETL workflows using DBT and Redshift DWH.
- Competitive Intelligence - crawl and analyze the data from competitors to drive insights using NLP.
- Analyzing feature usage data to understand customer behavior, predict conversion, and lead generation.
- Detecting product's commercial usage using machine learning to push for free-to-paid conversion.
- Generated country embeddings using relational data to identify geographic/demographic resemblance. 
- Clustering 200Mn+ users to tune marketing campaigns and generate sales opportunities.
- Conducting hiring interviews, providing mentorship and onboarding junior colleagues.

- Established a central framework and delivered end-to-end pipelines for 7+ ML projects into production.
- Recommender systems - Session-based (RNN), Collaborative Filtering (ALS) and Content-based.
- Product matching for 20Mn+ hotels using Fuzzy Pattern, NLP algorithms, and rule-based models.
- Time-series forecasting of hotel prices for next 90 days across 500+ cities using LSTM, ARIMA models.
- Customer journey mapping to understand user preferences before finalizing the purchase on website.
- Classify 13,000+ incoming mails every week using Text vectorization (NLP) and SGD classifier.
- Identifying 78% of the booking and voucher frauds using a random forest anomaly detection model.
- Created a RASA chatbot to help customers find answers to booking related questions and requests.

CHECK24 Vergleichsportal GmbH

- Deep Q-learning to train an autonomous car to navigate in an Industry 4.0 environment.
- Statistical and deep learning models for low latency (~3.8s) anomaly detection on acquired sensor data.
- Developed, tested, deployed, and managed continuous delivery and automation pipeline for this solution.
- Interactive visual explanations for results to understand, root-cause, and fix anomalous events.
- Deployed neural network models on resource-constrained devices using connection pruning strategies.
- Compressed computer vision deep learning models to 40% of original size with only a 6% accuracy drop.
- Presented technical papers to summarize the work and findings at the university and partner company.

Fraunhofer IEM, Paderborn, Germany

Research Scientist

- Developed emulation solutions - supporting 7+ clients, with a focus on system-level design / verification.
- Managed a team of 2 and involved in hiring resources, onboarding, and delivering training workshops.

Share