Data Engineering & Pipelines
Reliable dbt models and ETL/ELT workflows that turn raw, messy sources into clean, governed warehouse tables — because downstream models are only as good as the data feeding them.
• Data • Analytics • AI/ML
Data Professional
for messy, real-world data.
I turn messy data into pipelines, models, and dashboards people actually use — and I'm building toward AI/ML engineering.
- 3.8
- GPA · MS Data Analytics (Applied Machine Intelligence), Northeastern
- 3
- Industry certifications — Google, Tableau, Microsoft
- Dec 2026
- Expected MS graduation
- 12
- Open-source projects on GitHub
01 About
Hi, I’m Krunal.
I work at the intersection of data engineering, analytics and machine learning, building pipelines and tools that turn raw data into answers people can actually act on.
I started out in fintech, working on payments data at Webito, where a single duplicate transaction meant real money lost. Most recently I completed a Machine Learning Engineer co-op at METY Technology, and I’m finishing my Master’s in Data Analytics at Northeastern this December. Lately my focus is on where data meets AI: dbt pipelines, LLM-powered analysis tools and evaluating whether models actually get the right answer.
I’m looking for full-time roles across the data and AI stack, from building pipelines and dashboards to training and deploying ML models.
Krunal Vaghasiya
Data Professional · Data Engineering & Applied ML
Quick facts
- LocationBoston, MA
- RoleML Engineer (Co-op) @ METY Technology
- StackPython · SQL · dbt · Tableau
- FocusData engineering meets applied ML
02 Skills
The periodic table of my stack.
Twenty-three elements in four families. Hover or tap a tile to see how I use it — or pick a family to light it up.
What I do
Applied Machine Learning
From feature engineering to evaluation, I build and validate models with scikit-learn and XGBoost — and I'm honest about what the metrics actually say, including when a model isn't production-ready.
LLM-Powered Analytics
Retrieval-augmented tools and text-to-SQL interfaces built on APIs like Groq, then evaluated for whether the answers hold up — bridging the gap between LLM demos and dependable analytics.
Analytics & BI
Governed datasets turned into Tableau and Power BI dashboards that stakeholders actually open — clear visuals, honest numbers, decision-ready.
03 Approach Lab
How I turn messy data into decisions.
A practical view of the journey: raw data → cleaning & validation → pipelines & models → dashboards and decisions.
01
Messy data
Payments data • Sales exports • Survey files
02
Clean & validate
Small errors cost real money • Clean pipelines first
03
Pipeline & model
dbt • Warehouses • ML models that earn trust
04
Decisions & dashboards
Tableau • Power BI • LLM tools people use
04 Featured work
Things I’ve built.
Four projects that show how I think — from pipeline to model to the person reading the dashboard. Hover a panel to open it.
01
LLM AnalyticsFull-Stack
E-Commerce Product Analysis System
AI-powered strategic analytics platform — upload sales data, ask questions in plain English, and get actionable insights.
- Natural-language analysis over CSV, Excel, JSON and Parquet uploads
- Anomaly detection (Isolation Forest), sales forecasting, RFM segmentation, what-if simulation
- Groq Qwen 27B turns analysis into findings with real numbers
- FastAPI + Pandas backend, React frontend, Docker-ready
02
Data PipelinePublic Health
Opioid Pipeline Analysis
Multi-source analytics pipeline on the link between opioid prescribing, overdose mortality, and socioeconomic vulnerability across U.S. states.
- Prescribing rate alone doesn’t predict overdose deaths (R² = 0.043)
- Socioeconomic controls lift the model to R² = 0.369 — unemployment the top predictor
- Top-10 priority states ranked for intervention
- Interactive Tableau dashboard
03
ML EvaluationNLP
SMS Spam Detection
Feasibility study: can off-the-shelf ML services handle SMS spam detection? Tested on 5,572 messages.
- 93.54% accuracy — but 71.82% spam precision, below the 95% production bar
- Honest verdict: insufficient for production; AutoML recommended next
- Interactive web demo plus a 960-word APA research paper
04
RAGLLM App
AI Job Assistant
AI-powered assistant for job applications, built with retrieval-augmented generation over your own documents.
- RAG pipeline that grounds answers in resumes and job documents
- Assists with applications using your materials — not generic text
05 Experience
The journey.
Two stops, one through-line — turning messy data into pipelines, models and dashboards people can trust. Hover a card for a peek inside.
METY Technology
Machine Learning Engineer (Co-op)
- dbt pipelines and LLM-powered analysis tools
- Evaluating whether models actually get the right answer
Webito · Fintech / Payments
Data Analyst
- Payments data where small errors cost real money
- Clean pipelines before clever models
Education
MS, Data Analytics (Applied Machine Intelligence)
Northeastern University
BTech, Computer Engineering
LDRP University
Certifications
-
Google Data AnalyticsCoursera
-
Tableau Desktop SpecialistTableau
-
Microsoft Power BI AssociateMicrosoft
06 Projects
From the archive.
The rest of the archive — six more repos of experiments, pipelines, and tools, alongside the four featured projects above.
