Vishal Kumar
Analyst
Exploratory Analysis & Visualization: 7/10
Business Insight & Impact: 7/10
Data Wrangling & Cleaning: 6/10
Active 10 hours ago
+33 (7) 45497606 Invite to interview
Message
Download CVCV
Overview
Technical skills
Timeline
Roles
Overview
MSc student in Data Analytics & Artificial Intelligence at EDHEC Business School (Lille, France) and a BA (Hons)
Business Economics graduate from Delhi University. Experienced Research Associate at Jasper Colin, ranked
highest in the team for research accuracy in Dec 2024. Certified in Agile frameworks (PSPO I, Leading SAFe 6.0)
and proficient in Python, SQL, Power BI, and Tableau. Driven by a passion for turning data into strategic
decisions, with a growing global perspective shaped by international study in France.
Technical skills
AI/ML
Pandas
NumPy
Scikit-learn
Jupyter Notebook
Time Series Forecasting
Classic ML
Analytics
Matplotlib
Seaborn
Analytics
Timeline
EDHEC Business School Lille
Master's Degree •
Data Analytics and Artificial Intelligence
Research Associate
•
Middle
Jasper Colin
•
Full-Time
Assessed market characteristics, market potential, market share and sales performance of products and services across industries through primary and secondary research.
Researched and verified 200+ companies and contacts per week, building and enriching lead databases with LinkedIn and professional contact discovery tools.
Compiled 20+ structured research reports on consumer behaviour and market dynamics, organising quantitative and qualitative findings into standardised Excel templates for cross-functional teams.
Delivered survey, interview and secondary research that informed marketing tactics and strategy recommendations.
Ranked highest in the team for research accuracy (Dec 2024) through a two-source verification process and atraceable source log for every data point.
University of Delhi (DU)
Bachelor's Degree •
Business Economics
Middle Data Scientist
Confidence: Medium ML Practitioner
A practical mid-level ML practitioner specializing in applied forecasting and customer segmentation for retail use cases, with a strong strength in feature engineering and baseline-first evaluation.
Statistical Rigor
5/10
Correct use of statistics
Applied evaluation is present (appropriate metrics, baseline comparison and explicit correlation-vs-causation note) but formal statistical testing, uncertainty quantification and multiple-comparison controls are absent.
Evidence
Sales-Forecasting-with-Time-Series-Analysis/sales_forecasting.ipynb: 'Sales vs Oil Price' correlation cell and explicit 'correlation is not causation' note
Sales-Forecasting-with-Time-Series-Analysis/sales_forecasting.ipynb: rmsle / mape / evaluate function definitions and metric usage on test window
Data Wrangling & Cleaning
6/10
Preparing and cleaning data
Good practical data cleaning and leakage-aware feature engineering (lagging, rolling windows, interpolation); missing advanced provenance or automated data-validation tooling.
Evidence
Sales-Forecasting-with-Time-Series-Analysis/sales_forecasting.ipynb: oil interpolation and fill, lag_1/7/14 and rolling feature construction
Customer_Segmentation/Code.ipynb: dropna on CustomerID, removal of cancelled orders, removal of negative quantities and 99th-percentile outlier trimming in RFM
Exploratory Analysis & Visualization
7/10
Exploring and visualizing data
Exploratory analysis is purposeful and annotated; figures are saved and each visualization is accompanied by written interpretation tied to business questions.
Evidence
Sales-Forecasting-with-Time-Series-Analysis/sales_forecasting.ipynb: trend, seasonality and sales-vs-oil plots with inline interpretation and saved Figures
Customer_Segmentation/Code.ipynb: RFM distribution plots, elbow-curve and 3D cluster visualization with commentary
Predictive Modeling
6/10
Building models that predict
Solid predictive workflow with baseline-first discipline, time-based train/test split, feature engineering and model interpretation via feature importances; lacks cross-validation, systematic hyperparameter search and deeper error analysis.
Evidence
Sales-Forecasting-with-Time-Series-Analysis/sales_forecasting.ipynb: explicit Naive and Seasonal naive baselines, RandomForestRegressor training and evaluation on a final 90-day holdout
Sales-Forecasting-with-Time-Series-Analysis/sales_forecasting.ipynb: feature list and feature_importances extraction and plotting
Business Insight & Impact
7/10
Turning analysis into business value
Clear business framing and actionable recommendations with quantified impact estimates for forecasting and segmentation outcomes.
Evidence
Sales-Forecasting-with-Time-Series-Analysis/sales_forecasting.ipynb: 'Business takeaways' section describing operational actions and value of improved MAPE
Customer_Segmentation/Code.ipynb: 'Business Recommendations' section with segment-level strategies and revenue recovery estimates
Reproducibility & Notebook Hygiene
5/10
Clean, repeatable analysis
Notebooks include reproducibility basics (requirements.txt, fixed random_state, saved figures) and linear execution metadata, but there is no data versioning, CI, modularized pipeline code or environment pinning beyond a simple requirements file.
Evidence
Sales-Forecasting-with-Time-Series-Analysis/requirements.txt: lists pandas, numpy, matplotlib, scikit-learn, jupyter
Sales-Forecasting-with-Time-Series-Analysis/sales_forecasting.ipynb: RandomForestRegressor(random_state=42) and saved Figures; notebook meta shows linear execution
Industries
Commerce• Middle
Technologies
Analytics
Classic ML
Jupyter Notebook
Scikit-learn
Seaborn
Matplotlib
Pandas
NumPy
Time Series Forecasting
Python• mentioned only
Recommendations
- Build and productionize per-store, per-family forecasting pipelines that extend the existing one-step-ahead notebook into multi-horizon forecasts with windowed CV.
- Implement repeatable data pipelines (modular scripts or Airflow/Dagster jobs) to automate ingestion, validation, feature generation and model retraining for retail demand forecasting.
- Develop customer-segmentation tooling that operationalizes RFM clusters into marketing automation (exported segments, A/B test hooks and monitoring).
- Add systematic model selection and validation (time-series cross-validation, hyperparameter tuning, calibration and post-hoc error analysis) to move models from prototype to robust deployment.
Repositories
The developer's experience in this domain has been verified based on AI analysis of the following repositories: