I turn raw, messy data into decisions.
I'm Kalpesh Nankar, a Computer Engineering student at C.K. Pithawala College of Engineering & Technology (GTU), Surat — building Power BI dashboards, ML models, and end-to-end data pipelines from real, messy datasets.
Ingest
Merge messy CSVs & JSON APIs — TMDB, Spotify, flight fares.
Clean
Nulls, duplicates, outliers, nested JSON, categorical encoding.
Model
Regression, clustering, NLP, cosine similarity, RMSE / R² evaluation.
Visualize
Interactive Power BI dashboards & deployed Streamlit apps.
Projects
Six projects spanning NLP, regression, API pipelines, and BI dashboards — from raw data to deployed product.
- End-to-end content-based recommender on 4,806 movies — raw TMDB CSV to a deployed Streamlit app
- Merged two datasets, cleaned nulls, removed duplicates, parsed nested JSON for genres, cast, crew and keywords
- Built a unified tags column with Porter stemming and CountVectorizer (5,000 features), scored with cosine similarity across a 4,806 × 4,806 matrix
- Explored SVM, Naive Bayes, KMeans, hierarchical clustering (PCA-visualised) and Apriori association rules
- Responsive UI with parallel poster fetches via the OMDB API, hover popups and a 5-card grid across devices
- Cleaned an Indian domestic flight fare dataset — missing values, categorical encoding for airline, source, destination and stops, datetime feature extraction
- Full exploratory analysis: fare distribution, route analysis, correlation heatmaps, outlier detection
- Regression model to predict ticket prices, evaluated with RMSE and R²
- Interactive Power BI dashboard with a 30-day forecast, plus route, airline, and travel-date slicers
- Extracted and analysed track, artist and popularity data through the Spotify Web API
- Cleaned and transformed raw API JSON — normalised nested fields and handled missing values
- Surfaced patterns in music popularity and listening behaviour
- Interactive Power BI dashboard highlighting top-performing artists, genres and trends
- Multi-page Power BI dashboard for a fictional retail company, covering sales, customer and order analysis
- Built on retail transaction data with drill-through pages for order-level detail
- Designed for decision-support: revenue trends, customer segments, and order patterns in one workbook
- Processed raw Instagram follower data from OpenAI's account using pure Python — no external libraries
- Analysed post counts and follower-to-following ratios
- Classified accounts as Creator, Non-Profit or Personal and visualised the category distribution with Matplotlib
- Responsive Netflix landing page clone built with CSS flexbox and grid
- Semantic HTML, cross-browser compatible layout
Skills
Grouped the way I actually use them — from cleaning raw tables to shipping a dashboard.
Certifications
Data Analytics Job Simulation — Deloitte
Jan – Mar 2026Completed practical tasks in data analysis and forensic technology via Forage — simulated Deloitte analytics workflows including data interpretation, pattern analysis, and structured reporting.
Data Visualisation — Tata Group
Feb – Mar 2026Built executive-level dashboards and visualisations for business decision-making via Forage, analysing real datasets against Tata's industry reporting standards.
Currently learning
Why data.
I'm a Computer Engineering undergraduate specialising in data analytics, with a strong foundation in data science and AI/ML. I like taking a messy, real dataset and following it all the way through — cleaning, exploring, modelling, and finally turning it into something a decision-maker can actually use.
That's shown up as content-based recommenders, fare-prediction models, and Power BI dashboards built on API data, retail transactions, and streaming data. I've also completed job simulations with Deloitte and Tata Group to get a feel for how this work looks inside a real analytics team.
Right now I'm sharpening advanced SQL, statistics, and the fundamentals of deep learning — and looking for a Data Analyst / Data Science internship, on-site in Surat/Gujarat or remote.