Prompt search

Ask the system before the portfolio loads

>
Searching public portfolio graph
Reading ML, NLP, RAG, and forecasting systems
Matched profile: Data Scientist and AI Engineer
Opening Vatsal Rakholiya portfolio
Vatsal Rakholiya

Data scientist building predictive, NLP, and document intelligence systems.

Vatsal Rakholiya

Data Scientist | Intelligent Automation & Decision Analytics

Data scientist with production experience across Python, SQL, machine learning, NLP, forecasting, dashboards, and applied generative AI. I turn noisy business data into decision systems: lead scoring, support automation, market intelligence, document search, and visual inspection workflows.

Explore projectsContactLinkedIn
Toronto, Ontario, Canada
LinkedInGitHub

( The step aside )

AI should make operations clearer.

I turn scattered operational data into scoring systems, forecasting pipelines, NLP workflows, dashboards, and document intelligence products that teams can actually use.

Current focus

Industrial Image Processing Color Separation Automation

40%

manual support review reduced

16K+

crypto assets tracked

4K+

cities in forecasting pipeline

85%

forecasting accuracy reached

case files

Works

Good data systems communicate. Great models surprise.

Each project is framed like an operational model: inputs, decisions, outputs, and measurable business impact.

01

model system

Industrial Image Processing Color Separation Automation

Computer vision workflow for CMYK/TIFF prepress automation

Built a web workflow for CMYK/TIFF artwork separation with channel extraction, tolerance-based masking, region exclusion, grayscale previews, color inspection, and vectorized NumPy processing.

PythonNumPyOpenCVTIFFFlaskReact
  • - Reduced manual prepress review from 24+ hours to minutes
  • - Vectorized pixel-level processing

02

model system

Housing Operations Hub

Operations workspace for housing program visibility

Built a structured operations hub for housing teams to track program inputs, review status, reporting context, and decision-ready updates in one workflow.

Next.jsTypeScriptDashboardsWorkflow DesignData Quality
  • - Centralized housing updates, notes, and status tracking
  • - Improved review visibility across operational records
  • - Designed reporting views for faster stakeholder decisions
  • - Added quality checks for incomplete or inconsistent inputs

03

model system

Cryptocurrency Market Intelligence & Prediction Platform

Forecasting and sentiment intelligence across digital assets

Architected an analytics platform tracking 16,000+ cryptocurrencies through market APIs and automated ingestion pipelines, with price forecasting and real-time news sentiment.

TensorFlowLSTMGradient Boosted TreesBERTMarket APIsDashboards
  • - 16,000+ assets tracked
  • - Weekly percentage movement forecasting
  • - Real-time sentiment signals

04

model system

Document Intelligence System

Semantic search, summarization, and QA over private documents

Built document summarization, semantic search, sentiment extraction, and question-answering workflows using transformer embeddings, FAISS indexing, a Streamlit interface, and FastAPI backend services.

PythonHugging FaceFAISSFastAPIStreamlitDocker
  • - Transformer embeddings
  • - Retrieval-augmented QA
  • - Dockerized backend services

05

model system

Buyer Inquiry Intelligence Workflow

Lead scoring, NLP triage, and KPI operations

Built Python and SQL workflows for buyer inquiries, product requests, quotations, seller responses, follow-ups, and support notes.

PythonSQLScikit-learnNLPDashboards
  • - Developed lead-scoring models to identify higher-value buyer inquiries
  • - Created NLP workflows for intent, sentiment, topic grouping, and support ticket categorization
  • - Built KPI dashboards for inquiry volume, lead quality, follow-up status, response patterns, and support categories

06

model system

Commodity Price Forecasting Pipeline

Market intelligence across 4,000+ cities

Built a commodity price forecasting pipeline organizing market and location data from 4,000+ cities.

PythonSQLARIMASARIMARandom Forest
  • - Prepared SQL-ready commodity, city, and market datasets for reporting and model training
  • - Trained ARIMA, SARIMA, and Random Forest models to forecast steel prices and identify seasonal pricing patterns
  • - Achieved 85% forecasting accuracy and improved seasonality detection by 78% through tuning and feature analysis

research

Publications

Nov 2025

Adversarial Vulnerabilities in Machine Learning: Differential Analysis of Gradient and XGBoost-Based Attack Mechanisms

Paper ID: ICUIS-383

Compared gradient-based and XGBoost attack mechanisms and limitations of adversarial training for tree ensemble models.

capability map

Skills

Practical tools for machine learning, automation, analytics, and applied AI delivery.

01 / system

Programming

PythonSQLJavaScriptC++Base SASSAS Macros

02 / system

ML & AI

Predictive ModelingModel EvaluationTensorFlowPyTorchHugging Face Transformers

03 / system

Data Systems

PandasNumPySparkPostgreSQLNoSQLFAISSPineconeChroma DB

04 / system

NLP & GenAI

BERTBARTDistilBERTGPT-4LLaMAGeminiLangChainLlamaIndexRAG

05 / system

MLOps & Cloud

AWSEC2LambdaAzure MLGCP Vertex AIDockerMLflowCI/CDFastAPI

06 / system

Analytics & BI

TableauPower BIPlotlyStreamlitFlaskA/B TestingExperiment Design

credentials

Certifications

Coursera

Machine Learning Pipelines with Azure ML Studio

Built and deployed an end-to-end predictive pipeline using Azure ML Studio workflows.

Coursera

Agile Project: Product Prototype Touchpoint Analysis in Miro

Mapped product touchpoints in Miro to analyze user flow, feedback loops, and agile prototype decisions.

Coursera

Business Analysis & Process Management

Practiced process mapping, stakeholder analysis, and workflow improvement for business operations.

University of Pennsylvania via Coursera

Fundamentals of Quantitative Modeling

Studied model structure, assumptions, and quantitative reasoning for business and financial analysis.

Coursera

AWS S3 Basics

Configured S3 buckets, object access, encryption, versioning, lifecycle rules, and static hosting basics.

University of Michigan via Coursera

Understanding and Visualizing Data with Python

Used Python to summarize, visualize, and interpret univariate and multivariate data.

timeline

Experience

Apr 2023 - Jun 2024

Data Scientist

Multipz Technology / Ahmedabad, Gujarat, India

  • - Built Python and SQL workflows for buyer inquiries, product requests, quotations, seller responses, follow-ups, and support notes.
  • - Developed Scikit-learn lead-scoring models to help sales teams identify higher-value buyer inquiries.
  • - Created NLP workflows for intent classification, sentiment analysis, topic grouping, and support ticket categorization.
  • - Built KPI dashboards for inquiry volume, lead quality, follow-up status, response patterns, and support categories.

Jan 2023 - Apr 2023

Data Analyst Intern

Elogicals Technology / Ahmedabad, Gujarat, India

  • - Built a commodity price forecasting pipeline organizing market and location data from 4,000+ cities.
  • - Prepared SQL-ready commodity, city, and market datasets for reporting and model training.
  • - Trained ARIMA, SARIMA, and Random Forest models to forecast steel prices and identify seasonal pricing patterns.
  • - Achieved 85% forecasting accuracy and improved seasonality detection by 78% through tuning and feature analysis.

Education

Sep 2025 - Apr 2026

Post Graduate Diploma - Financial Technology

Seneca Polytechnic, Ontario, Canada / GPA 3.5/4.0

Sep 2024 - Apr 2025

Post Graduate Diploma - Artificial Intelligence

Fleming College, Ontario, Canada / GPA 3.8/4.0

Aug 2020 - May 2023

Bachelor of Engineering - Information and Communication Technology

Gujarat Technological University, Gujarat, India / GPA 3.9/4.0

contact

Send a project signal

Reach out about data science roles, ML systems, dashboards, document intelligence, or applied AI products.

rcm.vatsal@gmail.com