Apr 2020 – Nov 2021 · 1 yr 7 mos · Hyderabad, Telangana, India

• 1. Contextual Question Answering engine for HTA* reports.
• 2. Data Science engine for personalized tactic sequencing and content:
• AI/ML backed sales attribution from brick to HCP level.
• Comprehensive 360 view of the HCP along with micro clustering.
• Ensemble of predictive model and optimizer for personalized channel sequencing.
• Affinity models for personalized content.
• 3. CAPE(Capabilities For Analyzing Price Estimates)
• Innovative, integrated and interpretable tool built on python dash backed by Catboost and KNN models to provide Price range, Price and sales evolution of a prospective brand launch at Novartis.
• It does this by providing the top analogs for the to-be launched brand.
• I lead the conceptualization as well as development of the tool. Technologies Used - Python, Catboost, KNN, Plotly, Dash.
• 4. Dynamic Targeting - Identify the best promotional strategy for Novartis Oncology brands to dynamically target the stakeholders. Technologies Used - Python, Catboost, Hive, Impala.
• 5. Congress and Event Optimization: Optimize the congress and event invitations and the roles of the invitees using machine learning. Technologies Used - Excel, Decision Tree, Python.

Jan 2020 – Dec 2020 · 11 mos

• Authored and reviewed courses on Python and Machine Learning. Machine learning for finance is one of them with a rating of 4.5/5.

Jan 2020 – Dec 2020 · 11 mos

• Authored and reviewed courses on Machine Learning and Deep Learning.

Jul 2019 – Oct 2019 · 3 mos · Mumbai Metropolitan Region

• Helping students with doubt clearing sessions on NLP.

Feb 2019 – Apr 2020 · 1 yr 2 mos · Dubai, Dubai, United Arab Emirates

• UAE based Cinemas: Introduced new combos in the candybar menu to increase combo incidence by over 10 percent and revenue per admission by 4 AED. Techniques Used: Python, FP Growth, Apriori, A/B Testing(T-Test, Chi-Square), Power BI, SQL, Vertica.
• UAE based Cinemas: Developed a Customer 360 vision by introducing Segmentation, CLV and Churn models. Techniques Used: Python, R, Pareto/BG-NBD, LSTM, Logistic Regression, XGBoost, K- means, Tableau, SQL, Vertica.
• UAE based Leisure and Entertainment business: Revamped the pricing of different packages being offered to the customers in order to increase the NPS value and revenue. Techniques Used: Python, Simulated Annealing, Tableau
• UAE based retailer: Enabling the retail business by forecasting demand across 250 + stores across 2000 SKUs and decreasing the shortage by over 30 percent in 3 months of launch. Techniques Used: Python, LSTM, Prophet, DTW.
• U.S based investment bank: Build a knowledge graph and search mechanism to power the chatot. Techniques Used: Python, Spacy NER, Concept Extraction, Word2Vec and Neo4j.
• Sapient COE: Develop a generalized Media Mix Modelling accelerator which analyzes contribution, elastic, decay and carryover effect to recommend optimum budget allocation. Techniques Used: Python, Linear Regression, L1-L2, Elastic Net, Scipy Optimization.
• Sapient COE: Developed recurrent capsule networks for extractive summarization. Techniques Used: NLP, NLU, Deep Learning, LSTM, Capsule Networks, Machine Learning, Tensorflow, Python, Bleu Score, Rogue N Metrics, Tensorboard.

Jan 2019 – Sep 2019 · 8 mos · Gurugram, Haryana, India

• Mentoring students and industry professionals on chatbot and computer vision related problems including entity and intent extraction, semantic segmentation and image classification.

Jul 2017 – Feb 2019 · 1 yr 7 mos

• Client: Fortune 500 Retailer Of USA
• Procured the rating and review data across omni-channels and collating it with previous product demand for inventory forecasting across stores.
• Lead the team to build an end to end solution for categorizing and ranking the vendors to fulfill each shipment saving over 20 % in vendor default and shipment quality control issues.
• Technologies: Python, Collaborative Filtering, K-Mode clustering, XGBoost, ARIMA, Logistic Regression, Multinomial Naive Bayes, Feed Forward Neural Network, Keras, LSTM.
• Client: Major Manufacturing Company Of Germany
• Lead the effort to predict the future requirement of spare parts at various distribution centres of the client.
• Technologies: Python, RNN, LSTM, Keras, Pytorch, XGBoost, Time Series Forecasting, Pandas, Numpy, SGD, ARIMA.
• Client: Major publication of U.S.A
• Developed a text pruning solution to free up space on the front page of the newspaper by using unsupervised NLP techniques.
• Technologies: Python, NLP, Lex Rank, PyText rank, Gensim, Pandas, Scikit Learn, Bleu score, Rogue-N metrics.
• Client: Major Ecommerce player in India
• Developed a CNN based model using the Resnet 50 architecture to identify the label from the
• from product images.
• Technologies: Python, Deep Learning, Keras, SGDR, Transfer Learning, Computer Vision.

Jul 2015 – Jul 2017 · 2 yrs

• Electricity consumption intelligence for a Canada based client:
• Analysis of the electricity consumption data using linear regression in multiple variables.
• Predicting the future electricity consumption using devised model.
• Detecting anomalies in consumption pattern by using spring batch and K Means Clusteirng analysis.
• Key Responsibilities:
• Leading a team of 6 for end to end integration and delivery of the application.
• Conceiving design, frameworks and technology stack for the app.
• Developed the data analysis model for prediction of future electricity consumption.
• Technologies:
• R, Statistics, Machine Learning, Big Data, MongoDB, Spring Batch.
• Project: Gloria(https://www.globallogic.com/our-work/gloria-chatbot/)
• Developed a closed domain voice driven chatbot paired with IoT system to control the devices around.
• Technologies: RASA, Bluetooth Low Energy module, Arduino, Spring Boot, Gradle.
• Wallet Application
• Wrote Splunk queries and spring batch jobs to generate descriptive analysis insights from streaming log files.
• Improved the accuracy of data access layer by 20 to 30 percent by caching the retrieval of more than 20 million data objects nearing 20 GB.
• Developed and designed highly scalable and high throughput offer prediction model using xgboost classifier.
• Single handedly developed a tool to schedule periodic health checks of the REST services.
• Chat Application for the support associates in the retail store
• Technologies Used: Spring Boot, Gradle, REST, Mosquitto Broker, Cassandra database.
• Responsibility: Creating the RESTful web services to be consumed by the android and ios chat client. Configuring the MQTT broker to persist the messages into cassandra.

Aug 2014 – Jul 2015 · 11 mos · Gurugram, Haryana, India

• AP Moller - Maersk
• Optimisation of possible routes for evacuation of containers from a surplus
• site to a deficit one using SAS. Showing these routes on google maps.
• Cleaning up of huge csv data and load into the oracle database by writing ETL jobs.