DriverAnshav Desai TeamData science / ML Clock (GMT+5:30)--:--:--.---

Data science & machine learningMultiple Karting Championship WinsFounder of IPR Racing Team

I build machine learning and data systems, and also race karts. Both come down to the same job: get clean data, find where the time is going and fix it.

Long-term target: performance and strategy analysis.

Live telemetry · this visitRec
CH1 pointer0px/s
CH2 scroll0px/s
Real numbers from your pointer and scrolling, sampled 30 times a second over the last 8 seconds. Nothing leaves your browser.
9.61CGPA · BBA (CA), SPPU
98.6%Accuracy · bird vs drone
≈0.94F1 score · 10-class food model
5Championship Wins · 2024 → 2026
Sector 1

Profile

Nashik, India

I finished a Bachelor of Business Administration with Computer Application at Savitribai Phule Pune University in 2025 with a 9.61 CGPA, then went all in on Data Science.

That meant the Data Science Mastery Program by IIT Madras Pravartak with GUVI and HCL, Imperial College London's mathematics for machine learning courses and two internships. At Siam Computing I worked on a unified healthcare platform: five separately built projects merged into one backend, one frontend and one role-based login.

I learn by shipping things: computer vision pipelines, fine-tuned CNNs, an analytical SQL database. Each one is linked below with its code.

Off the screen I race rental karts and run "It's Part of Racing", a karting team I founded. One of its formats is a lap-time Consistency Championship, which is really a statistics problem wearing a helmet.

Toolkit, and where it was used
AreaToolsProject
Computer visionOpenCV, SIFT, RANSAC homography, morphology001
Deep learningPyTorch, TorchScript, Grad-CAM002
Deep learningTensorFlow, Keras003
Transfer learningResNet-50, ResNet-101, EfficientNet-B0002003
DeploymentStreamlit003
Data & BIPython ETL, SQLite, SQL, Power BI004
LLM appsRAG, LangGraph, Gemini APIInternship
AlsoPandas, NumPy, scikit-learn, Matplotlib, Seaborn, Tableau, Git
Sector 2

Selected work

4 projects · newest first
001RGB-Thermal Alignment & Change DetectionTwo computer vision pipelines in Python and OpenCV2 pipelinesFully batch-automatedDec 25 → Jan 26

Two separate problems, both solved as automated batch pipelines that run over whole folders of images.

  • RGB to thermal alignment: SIFT features and a RANSAC-estimated homography warp each thermal image onto its RGB pair.
  • Scene change detection: image differencing, then morphological processing and connected-component analysis to isolate objects that went missing between frames.

Tech stack

  • Python
  • OpenCV
  • NumPy
  • SIFT
  • RANSAC
  • Morphology
  • Connected components
Repository
002Food Image Classification10-class classifier, three architectures compared≈0.94F1 scoreNov 25 → Dec 25

A full PyTorch pipeline, from data splits to an exported model.

  • Stratified train, validation and test splits, image augmentation and ImageNet normalisation.
  • Fine-tuned ResNet-50, ResNet-101 and EfficientNet-B0 with AdamW and a cosine learning-rate schedule, then compared them.
  • Checked what the models actually learned with confusion matrices and Grad-CAM heatmaps.
  • Exported the best model with TorchScript.

Tech stack

  • PyTorch
  • Transfer learning
  • CNN
  • AdamW
  • Grad-CAM
  • TorchScript
Repository
003Aerial Object ClassificationBird or drone, with a live web app98.6%AccuracyOct 25 → Nov 25

An end-to-end binary image classifier, built twice and compared: a custom CNN against a fine-tuned EfficientNetB0.

  • Handled class imbalance in the training data.
  • Built in TensorFlow and Keras, then fine-tuned the transfer-learning model.
  • Deployed as a cloud-hosted Streamlit app that returns a prediction with its confidence score.

Tech stack

  • TensorFlow
  • Keras
  • CNN
  • EfficientNetB0
  • Streamlit
  • Deployment
Repository
004Nutrition Paradox: Global ViewObesity and malnutrition, side by side, from WHO data25+SQL queriesAug 25 → Sep 25

Why do some countries face rising obesity and persistent malnutrition at the same time? This project builds the data to look at that question.

  • ETL and feature engineering on WHO nutrition datasets.
  • A SQLite analytical database with 25+ queries for trends over time and across demographics.
  • Power BI dashboards that put the two sides of the paradox next to each other.

Tech stack

  • Python
  • ETL
  • SQLite
  • SQL
  • Power BI
  • WHO data
Repository
005

In the garage: an open Formula 1 data site with a race archive, telemetry views, driver and team comparisons, and lap-time predictors. Early stage.

Building
Sector 3

Experience

Newest first

Internships 2

Apr 26 → Aug 26

AI/ML Intern · Siam Computing

Three solo builds: an F1 RAG chatbot, an AI patient-intake system that writes triage reports, and an insurance navigator. The internship was then extended so I could merge five healthcare AI modules, built separately by different interns, into one platform with a single backend, a single frontend and one role-based login for doctors and patients. Free-tier APIs only.

4 months
Feb 26 → Mar 26

Machine Learning Intern · MyDailyWork

Built machine learning pipelines for a credit card fraud detection system, a customer churn predictor and a text generator.

1 month

Education 2

Apr 25 → Nov 25

Advanced Programming Professional: Data Science Mastery · GUVI HCL & IIT-M Pravartak

A rigorous, six-month professional certification course with six data science capstone projects.

95.94% in Final Test
Jul 22 → May 25

Bachelor of Business Administration with Computer Application · Savitribai Phule Pune University

College: Ashoka Centre for Business and Computer Studies, Nashik (affiliated with Savitribai Phule Pune University).
Class Representative, and selected for the college's Elite Club for academic excellence and initiative for 2 consecutive years.

CGPA 9.61

Courses & certifications 8

Coursera

Data Science Math Skills · Duke University

In Progress
Certificate

AI Fluency: Framework & Foundations · Anthropic

September 2026
Coursera

Mathematics for Machine Learning · Imperial College London

Multivariate Calculus.

April 2026
Coursera

Mathematics for Machine Learning · Imperial College London

Linear Algebra.

March 2026
Certificate

Artificial Intelligence · MSME-PPDC

August 2025
Certificate

Artificial Intelligence · GUVI with HCL, at Techfest, IIT Bombay

December 2024
Course

IoT using Arduino · Cognifront

September 2024
Course

Machine Learning · AcmeGrade with Mood Indigo, IIT Bombay

August 2023

Outside work 2

2 years

Class Representative · Ashoka Centre for Business and Computer Studies

Acted as a liaison between the student body and the faculty, and represented student interests.

5 years

Volunteer fundraiser · HelpAge India

Raised funds for elderly people across India by going door-to-door to spread awareness about the cause.

Sector 4

On track

Results 2024 → 2026

Competitive rental go-karting across Western & Central India, solo and in team endurance races.

My team, the "It's Part of Racing" Racing Team (IPR Racing), is a team of racing enthusiasts from several Indian cities that runs its own formats, including a lap-time Consistency Championship, hosts motorsport-related discussions and debates, and shares a platform for discovering and discussing upcoming rental karting events across India with other team members.

Apply to join IPR Racing
5Wins
8Podiums
11Total Competitions
Karting results, newest first
YearEventVenueResult
2026Mighty 120 Endurance Team RacePuneP2Vice Champions
2026Trackfest, The Grand Prix Finale ChampionshipMumbaiP1Champion
2026Trackfest, The Endurance Team Finale ChampionshipMumbaiP1Champions
2026Grand Prix ChampionshipIndoreP1Champion
2026Endurance Team ChampionshipVirarP1Champions
2025Red Rabbit Racers Talent HuntMumbaiP2Vice champion
2025Red Rabbit Racers Talent HuntPuneMedalRising Star
2024Grand Prix ChampionshipVirarP2Runner-up
2024Grand Prix ChampionshipNashikP1Champion

Win: purple, like the fastest sector on a timing screenPodiumAward

Reaction test

Five red lights, then lights out. Tap the gantry (or press Space) to arm it, then react the moment the lights go out. Go early and it is a jump start. Do five and the panel beside it scores how consistent you were.

Stint analysis · last 5 starts
    Stint best
    -.---
    Average
    -.---
    Spread
    -- ms
    Device best
    -.---
    Spread is the standard deviation of your clean starts. Lower means more repeatable, which is what a consistency championship rewards on track.

    Times include your screen and input lag, so treat them as a fun number rather than a lab result.

    Lap Complete

    Contact

    IST · GMT+5:30

    Open to freelance projects, collaborations and 3 to 4 month internships. Data, analytics, machine learning and anything with a lap time attached.

    Status: AvailableBased in: Nashik, India