Ashish K
Data Scientist
PROFILE SUMMARY
Demonstrated sound business judgement, well- developed planning, analytical and communication skills and a consistently high level of performance in a variety of progressively responsible and challenging roles. I have provided solutions and consultation to mid and large clients. I am accustomed to a fast-paced environment where deadlines are a priority and can handle multiple jobs, simultaneously. I have a strong desire to learn and seek out new relevant technologies.
WORK EXPERIENCE AND INTERNSHIPS
Data Scientist
Marktine Technology Solutions Pvt. Ltd. Jaipur, Rajasthan, IN
12/21 – Present
Junior Data Scientist
TechnoStruct Gurgaon, Haryana, IN
04/17 – 11/21
Project: Precast Structure Compressive Strength.
- Proposed Random Forest regressor model to predict the erected precast structures' compressive strength. Able to achieve RMSE 7.34.
- Performed data preprocessing techniques. Used KNN imputer to handle missing values. Performed feature engineering and scaling on various features.
- Experimented with various algorithm like Linear Regressor, Decision Tree, Support Vector, Naïve Bayes.
- Model is in testing phase with our trusted clients. The proposed goal is to reduce time of concrete manufacturing by 17-22%.
Project: Sales Insight Dashboard For MEP Equipment.
- Built a sales insight dashboard using Tableau which proved 10% cost saving on total spent. Saved 20% time of sales analysts by stopping manual data gathering.
- Used AIMS Grid for project planning and defining purpose, stakeholders, success criteria, final result. Coordinated among stakeholders through RFI and Kick Off.
- Performed ETL using MySQL and Tableau. Used star schema to define relations among customer, data, markets, product and transactions tables from Sales database.
- Employed feature engineering on transaction table by adding new feature like normalized amount, profit and profit margin.
Project: Lead Qualifier
- Identified the problem of leads being not converted to revenue. Bult a model that qualified leads to revenue. Performed deep analysis on company's leads data.
- Involved in preprocessing, feature engineering, EDA and model building.
- Built a XG Boost model with precision score of 0.72. Experimented with other ML models.
- leads to revenue conversion increased by 11%. Helped marketing team to run campaign efficiently on target customers.
Intern (Data Science)
Fliprobo Technologies Bangalore, Karnataka, 2020
Project: Wholesale Store Sales Prediction Created a new Deep Learning based Hybrid model for Stock Prices prediction.
- Developed a stacking model to help Store Managers predict their future sales and be able to take necessary upstream & downstream decisions to optimize for supply chain cost and customer experience.
- Employed essential data pre-processing techniques. Analyzed pattern of store sales to capture the cyclic trend over different time period, relation with marketing campaigns, store demographics and nearby competitors.
- Engineered features like- 'Total Competition Month, Total Promotion Year, Total Promotion Week, IsPromoMonth, Average Sales and Average Customers. Used One Hot Encoding to treat categorical features.
- Used naïve forecasting as a benchmark model with 33% MAPE. Implemented a stacking model of Decision Tree Regressor, Random Forest Regressor and Light GBM to reduce MAPE to 18%. Used feature selection techniques to get important features.
Project: Sentiment Analysis of Tweets to Flag Negative Sentiments.
- Built a binary-class Naive Bayes classifier to predict the sentiment associated with a particular tweet and obtained the F1 score of 0.79.
- Employed essential text preprocessing techniques such as tokenization, regular expression, stop words removal and stemming to obtain clean tweets. Used Word Cloud to visualize and analyze the tweets.
- Performed feature engineering on Sentiments and converted it into binary classification. Used word embedding and generated optimal document-word sparse matrix using TF-IDF Vectorizer by tuning parameters.
- Experimented with multiple classification algorithms such as Multinomial Naive Bayes, Random Forest, XGBoost and SVC to compare with binary classification model.
Intern (Data Science)
AlmaBetter Bangalore, Karnataka,
2021
Project: Book Recommendation System.
- Developed a book recommendation system for customers using memory- based collaborative filtering by utilizing the description of book and user interests.
- Engineered new feature- 'score to check the popularity and employed cosine distance to measure user-term and item-term similarities.
- Created profiles for top active users by leveraging interaction strength with the recommended items and achieve test recall@5 of 42% and recall@10 of 53%.
- Analyzed the solution for the cold start problem based on global and demographic specific book popularity and improved efficiency of the user recommendation engine.
TECHNICAL SKILLS
Languages: SQL, Python
Frameworks and Libraries: Keras, Tensorflow, OpenCV, Pandas, NumPY, Scikit-Learn, NLTK, Matplotlib
SPECIALIZATION AND COURSES
- Udemy- The Complete Python Course
- Udemy-SQL & PostgreSQL for Beginners
- Udemy- The Complete Machine Learning Course
- Udemy- Feature Engineering for Machine Learning
HOBBIES AND PERSONAL INTERESTS
Swimming, Reading, Horse riding, Traveling