Summary
Overview
Work History
Education
Skills
Hobbies and Interests
Projects
References
Timeline
Generic

Shitij Arora

Delhi

Summary

Seasoned Data Scientist with a proven track record at UCBOS, enhancing computer vision, natural language processing, predictive analytics and operational efficiency by applying advanced statistical analysis and state of the art model development. Spearheaded innovative AI solutions, achieving a significant increase in the output. Excelled in leadership, guiding teams towards groundbreaking achievements with a keen focus on leveraging data for strategic insights.

Overview

7
7
years of professional experience

Work History

Senior Data Scientist

UCBOS
Bangalore
09.2021 - Current
  • Scheduled weekly updates for asset monitoring using advanced machine learning techniques.
  • Integrated GPT-3, LangChain, and Transformers in developing an automated conversational tool to streamline query resolution related to the UCBOS application.
  • Integrated OCR within object detection to leverage the detections from the model, and using OCR to extract the text. Used this solution as a record generator, and multiple other solutions.
  • Analyzed data from various sources to generate insights into customer behavior and segmentation.
  • Created an AI-driven customer service chatbot based on GPT-4, LangChain, and Transformers.
  • Developed and implemented a dynamic slotting solution to optimize warehouse storage, reducing storage and labor costs by 15%. Utilized SQL, Pandas, NumPy, and TensorFlow for the solution.
  • Developed a solution for scheduler-driven training of the model with new input in the data, with planned and forecast quantities, which allowed the retail facilities to maintain a required stock for each week.
  • Implemented regression and classification models using PySpark, achieving a 90%+ accuracy rate for predictive analytics. Applied these models to customer churn prediction, increasing retention by 25%.
  • Developed LSTM models with TensorFlow for time series analysis, improving forecast accuracy by 20%. Provided solutions for lead time insights, enhancing supply chain efficiency.
  • Designed SHAP-driven explainable AI models for neural networks and PySpark, enhancing interpretability by 30%. Helped clients understand predictions and feature importance, boosting decision-making efficiency.
  • Applied ARIMA, ARIMAX, and Holt's Winter models for sales forecasting, enhancing accuracy to 92%. Used these models to optimize inventory levels, and reduce wastage by 18%.
  • Developed in-memory dynamic object detection and image classification solutions with OpenCV and TensorFlow/PyTorch. Implemented for retail shelf replenishment, increasing restocking efficiency by 25%, and for Pallet Detection and Count to verify damage.
  • Utilized multiple survival analysis models to enhance cancer survival analysis, resulting in a 15% improvement in prediction accuracy.
  • Achieved a 90% accuracy rate by applying BERT-based sentiment analysis to enhance customer feedback insights.
  • Applied panel data modeling to corporate finance issues, improving financial forecasting precision.
  • Supervised a team of four to handle project challenges and achieve alignment among stakeholders.
  • Engineered AI/ML layer from ground up, facilitating integration with ActiveMQ and MarkLogic database. Developed comprehensive Python code for the platform's PaaS functionality.
  • Engineered REST interfaces to facilitate model executions and integrate custom coding processes using Flask for data management.

Data Scientist

Arthashastra Intelligence
New Delhi
03.2021 - 09.2021
  • Analyzed target audiences using Random Forest, Decision Trees, and XGBoost, increasing campaign effectiveness
  • Developed time series models to forecast sales with LGBM, reducing inventory costs by 15%
  • Built a movie recommendation system app using Flask and AWS EC2, improving user engagement
  • Created a predictive model for house prices, achieving a 95% accuracy rate
  • Utilized BERT and GPT-2 for advanced NLP models, enhancing topic clustering accuracy by 25%

Business Development Manager

Acer India
New Delhi
09.2018 - 04.2020
  • Identified potential clients and established relationships

Education

Bachelor of Technology - Electronics & Communication

GGSIPU
Delhi
12-2018

High School Diploma -

DPS Rohini
Delhi
04-2014

Skills

  • Python Development
  • Machine learning
  • Time series forecasting
  • Anomaly detection
  • Predictive analytics
  • SQL
  • PyTorch
  • TensorFlow
  • NumPy
  • SciPy
  • OpenAI Gym
  • Pandas
  • PySpark
  • Transformers
  • Experience with multiple cloud services

Hobbies and Interests

  • Football
  • Playing guitar
  • Hiking

Projects

Tuning LLM for personal data: Utilized Llama and GPT-3 

Image Captioning: Utilized Neural Networks and Transfer Learning with Resnet50, achieving a 90% accuracy rate in image captioning. 

Sentiment Analysis on IMDb Movie Reviews: Developed a 4-layer MLP for sentiment analysis. 

Telegram News Bot: Created a news-fetching Telegram bot using gnewsclient in Python. 

Movie Recommendation System: Built a system based on IMDb reviews data. 

Face Recognition System: Implemented face recognition with OpenCV2 and Haar cascade filters. 

Titanic Survivor Prediction: Predicted survival chances with Decision Trees and Random Forest.

 Image Classification: Classified images into categories (dog, cat, horse, human) using SVMs.

References

References available upon request.

Timeline

Senior Data Scientist

UCBOS
09.2021 - Current

Data Scientist

Arthashastra Intelligence
03.2021 - 09.2021

Business Development Manager

Acer India
09.2018 - 04.2020

Bachelor of Technology - Electronics & Communication

GGSIPU

High School Diploma -

DPS Rohini
Shitij Arora