Spandita Biswas
Bengaluru
Summary
Data Analytics and Engineering professional with 2 years of experience in designing and implementing end-to-end data solutions and scalable ETL pipelines. Expertise in developing BI reporting systems that facilitate data-driven decision-making. Demonstrated success in automating operations to enhance efficiency and achieve measurable results. Proven track record of leading teams to complete projects on time and within budget.
Overview
1
1
year of professional experience
2
2
Certifications
Work History
Decision Scientist
Mu Sigma
Bengaluru
08.2024 - Current
- Store Audit Analytics: Sales representative seeks to capture shelf images in stores using a mobile app and instantly generate automated analytical reports.
- Implemented backend infrastructure and developed PySpark scripts on Databricks to ensure KPI compliance for SKUs, resulting in successful product deployment across 7+ countries.
- Supply Chain & BI Migration : Orchestrated the migration of legacy QlikView dashboards to a modern Power BI stack, refactoring static Excel-based analyses into dynamic, automated data pipelines.
- Extracted data from diverse sources including Excel files, CSVs, and SQL servers; built robust Azure Data Factory (ADF) pipelines to ingest and stage data into the Azure Data Lake (Delta format).
- Developed PySpark transformation logic to clean, join, and enrich datasets, incorporating business logic for key supply chain metrics such as GSV (Gross Sales Value) and NSV (Net Sales Value).
- Partnered with business stakeholders to translate supply chain KPIs into scalable data assets, leading to a 90% reduction in manual effort and significantly improved reporting reliability.
- Material Requirement Planning (MRP) & Data Engineering: Directed end-to-end delivery of an inventory management data solution: from SOW, requirement gathering to development, testing, deployment, and production sign-off.
- Built complex PySpark transformation logic to compute material requirements based on configurable lead times, current inventory levels, and consumption trends.
- Persisted transformed outputs to 21+ inventory-related KPIs into Hive (Delta/Parquet) tables for scalable querying and downstream consumption.
Education
Bachelor of Technology - Electronics And Telecommunications Engineering
Kalinga Institute of Industrial Technology
Bhubaneswar09-2024
Skills
- Data engineering
- Data Analysis
- Azure Data lake
- Project Management
- PySpark
- Stakeholder Engagement
- Python
- SQL
- Azure Data Factory
- Spark
- Dashboard Visualization
- ETL
- DevOps
- Git
Projects
1. SCALABLE CUSTOMER MARS OPERATING SYSTEM(SCMOS):
Client : MARS Incorporated
- Served as the primary point of contact for the legacy product across multiple markets, ensuring seamless coordination between SCA and DevOps teams.
- Managed and responded to product-related inquiries, successfully resolving 73 backend production issues, while fostering strong collaboration across cross-functional teams.
- Led rapid product iterations based on market needs, implementing key enhancements that improved product usability from 36% to 82%
2. Building a CNN Model with SVHN Dataset:
In this project I tried to create a basic CNN (Convolutional Neural Network) model for SVHN (Street View House Numbers) dataset and tried to achieve maximum accuracy with this basic CNN model. I chose SVHN as it is a real world data set of house numbers from google street view images. SVHN consists of 6,00,000 images. There are 2 types of data in SVHN, one is full numbers which are original image with character level bounding box and other is cropped image which are single character with distractors. And I used cropped images as it is easier to train ML model, increase accuracy and avoid irrelevant information. CNN is network architecture for deep learning algorithm and specially used for image recognition as well as it is highly suitable for computer vision task. Along with all this CNN can automatically learn to extract features from images. We used Relu activation function in hidden layers to achieve non-linearity and softmax at last to predict class of input image. And after 50 epochs I got accuracy of 90.7% for our model.
Languages
English
Proficient (C2)
C2
Hindi
Proficient (C2)
C2
Bengali
Proficient (C2)
C2
Certification
Foundations of Coding Back-End, Coursera, 2025
Affiliations
Volunteer, National Service Scheme
Timeline
Decision Scientist
Mu Sigma
08.2024 - Current
Bachelor of Technology - Electronics And Telecommunications Engineering
Kalinga Institute of Industrial Technology