Skip to content

B.E. Electronics & Computer Engineering · Thapar Institute · Class of 2027

Madhav
Arora

I build systems that make separate parts work as one: images and text in a single model, 140 hospitals in one federated study, three vendor systems in one app.

See projectsDownload resumePDF

Open to SDE, AI/ML, data and business analyst roles. Jump to the work that fits:

Pixels + words140 hospitals3 vendor systemsOne systemU1

Thapar Institute · CGPA 8.04/10

About

I’m a final-year Electronics & Computer Engineering student at Thapar Institute in Patiala. Most of my research sits where different kinds of data meet — images and text, clinical records spread across 140 hospitals — and on the engineering it takes to make that work outside a notebook.

Alongside research as a University Research Fellow, I build software people use: a live website for a printing press, and the MVP of Little’Learners, an assistive learning platform for children with dyslexia, dyscalculia and dysgraphia, where I lead technical development.

Since 2023 I’ve also been on the organising side of campus: mentoring 300+ students and managing budgets for festivals with 10,000+ participants.

Education

Thapar Institute of Engineering & Technology

B.E. Electronics & Computer Engineering · 2023 – 2027

8.04 / 10CGPA

Coursework

Data Structures & Algorithms · Operating Systems · DBMS · Computer Networks · Computer Architecture · Control Theory · Cloud Computing · Data Science & AI · AI/ML & Applications

  • Class XII (CBSE)Kundan Vidya Mandir, Ludhiana, 202389.6%
  • Class X (CBSE)BCM Sr. Sec. School, Ludhiana, 202192.4%

Recognition

  • University Research Fellowship (URF)Awardee, FY 2025–26
  • InovateX — 1st placeLittle’Learners AI platform
  • Create’n’Conquer — 3rd placeEgocentric data pipeline
  • International English OlympiadCity Gold Medallist, 2018

Selected work

Projects

Research, products and client work. Each one is tagged by track, so you can filter for the role you’re hiring for. The strongest project in each track is written up in full.

9 projects · 4 tracks

Federated ICU mortality prediction

  • AI/ML
  • Data
Context
Research internship · Washington State University
Period
Sep 2025 – Mar 2026
Status
Paper under review · IEEE Trans. Dependable & Secure Computing

A multimodal federated learning pipeline for predicting ICU mortality on the eICU dataset, covering 10,000 patients from 140 hospitals.

  • Designed FixedFusionNet, a three-branch model (BiLSTM + TF-IDF + note-type) that keeps any one modality from dominating — the “modality collapse” problem in distributed training.
  • Proposed a local weighted BCE correction that cut per-hospital variance by 32.4% compared with FedAvg.
  • Built and version-controlled the pipeline in Git, refining it through twice-weekly reviews with two mentors.
AUROC · FixedFusionNet + FedProx
0.7406
deaths detected
120/217
per-hospital variance vs. FedAvg
−32.4%

Built with: Federated learning / FedProx / FedAvg / BiLSTM / TF-IDF / Git

Little’Learners

  • Business
  • SDE
  • AI/ML
Context
AI Engineer & Head — Planning and Outreach · Start-up
Period
Dec 2025 – Present
Status
MVP in development · 1st place, InovateX

An assistive web and mobile learning platform for children aged 4–12 with dyslexia, dyscalculia and dysgraphia, using computer vision, NLP and AR to power therapeutic mini-games.

  • Lead technical development and product planning; architecting the MVP’s database, admin dashboard, and tracking portals for parents, teachers and clinicians.
  • Review code and coordinate a cross-functional team of engineers and paediatric specialists across India, California and Zurich in an Agile workflow.
  • Run an efficacy research and survey track alongside development to validate the mini-game interventions.

Built with: Computer vision / NLP / AR / Agile

PSPCL utility platform

  • SDE
  • Data
Context
Capstone project · Punjab State Power Corporation Ltd.
Status
In progress

One application that brings together billing, fraud detection and OCR/YOLO model pipelines, which today are split across three vendor systems.

  • Engineering a single platform with a unified database, dashboard and user interface to replace three separate vendor systems.
  • Integrating the OCR and YOLO model pipelines into the same application as billing and fraud detection.

Built with: OCR / YOLO / Unified database / Dashboard

More work

  • Pixels Meet Words & CrisisStream

    First-author research · University Research Fellowship

    ViT-B16 + BERT fusion for classifying informative crisis posts on CrisisMMD — 85% accuracy (F1 0.81, AUC 0.892) — extended into CrisisStream, a real-time triage pipeline with ≤2 s latency.

    • AI/ML

    Accepted, I2ITCON 2026 · Abstract, TICSR’26

    Built with: PyTorch / ViT-B16 / BERT / CrisisMMD

  • Attacks on invisible watermarks

    ELC Research Activity · Thapar Institute

    Designed and evaluated black-box adversarial attacks against four benchmark invisible-watermarking systems — StegaStamp, InvisMark, TreeRing and StableSignature — reaching a 71.4–79.86% removal rate.

    • AI/ML

    Research

    Built with: GANs / CNNs / Adversarial ML

  • Modern Monopoly

    Personal project

    A business-simulation game on the MERN stack with a reinforcement-learning bot opponent and dynamic market events.

    • SDE
    • AI/ML

    Core game logic done · bot and UI in progress

    Built with: MongoDB / Express / React / Node.js / Reinforcement learning

  • Dharamyug Press

    Client website

    Designed and deployed the website for a printing press business in Jalandhar.

    • SDE

    Live

    Built with: Web design / Vercel

  • VK Coaching Centre

    Website

    Website for a coaching institute in Ludhiana.

    • SDE

    Live

    Built with: HTML / CSS / Vercel

  • Egg-O

    Hackathon · Create’n’Conquer

    An egocentric data pipeline built for the Create’n’Conquer hackathon.

    • Data
    • AI/ML

    3rd place

    Built with: Data pipeline

Research, roles & papers

Experience

Roles

  1. Dec 2025 – Present

    AI Engineer & Head — Planning and Outreach

    Little’Learners (start-up)

    Leading technical development and product planning for an assistive learning platform; architecting the MVP and coordinating a team across India, California and Zurich.

    Read the project write-up ↑

  2. Sep 2025 – Mar 2026

    Research Intern — Federated Learning

    Washington State University

    Multimodal federated learning for ICU mortality prediction across 140 hospitals. Best AUROC 0.7406; paper under review at IEEE Trans. Dependable & Secure Computing.

    Read the project write-up ↑

  3. Summer 2025

    ELC Research Activity — Invisible Watermarking Attacks

    Thapar Institute of Engineering & Technology

    Black-box adversarial attacks (GANs, CNNs) against four benchmark watermarking systems, with a 71.4–79.86% removal rate.

    Read the project write-up ↑

Publications (6)

  • 1st authorI2ITCON 2026Accepted

    Pixels Meet Words

    ViT-B16 + BERT fusion for CrisisMMD informativeness classification; 85% accuracy, F1 0.81, AUC 0.892.

  • 1st authorTICSR’26Abstract

    CrisisStream

    Real-time multimodal triage pipeline on CrisisMMD with ≤2 s latency.

  • 1st authorTHEC 2025Presented

    MOSAIC

    A cultural-knowledge-graph-guided multimodal fusion framework.

  • 2nd authorTHEC 2025Presented

    From Classroom to Cubicle

    Mixed-methods study of internship impact, introducing the CCTM career-growth model.

  • 3rd authorCVIP 2025Conference paper

    DisasterNet

    TPF-MBT fusion for CrisisMMD damage-severity classification; 80.95% accuracy, 8+ points above prior state of the art.

  • 2nd authorAIMLA 2025Published

    Enhancing Strabismus Diagnosis from Detection to Classification with Deep Learning

    Self-acquired dataset; benchmarked EfficientNet, ResNet and VGGNet, reaching 92% accuracy.

Leadership

  1. 2023 – Present

    Official Mentor & Society Representative (FROSH)

    Thapar Institute of Engineering & Technology

    Mentored 300+ students through technical workshops and one-on-ones; coordinated logistics and content across campus societies for events reaching 4,000+ people.

  2. 2023 – Present

    Executive Committee Member — Saturnalia & URJA

    Thapar Institute annual festivals

    Managed Rs 2.5 lakh+ budgets and led marketing and outreach across 4 large-scale events with 10,000+ participants, growing reach by more than 10%.

Grouped by what they’re for

Skills

Languages
  • Python
  • C/C++
  • JavaScript
  • SQL
  • HTML/CSS
ML frameworks
  • PyTorch
  • TensorFlow
  • Keras
  • Scikit-Learn
  • OpenCV
  • YOLO
Models & methods
  • CNNs
  • RNNs
  • GANs
  • Transformers
  • LLMs
  • Federated learning
  • Prompt engineering
Data & analysis
  • Pandas
  • NumPy
  • Matplotlib
  • Seaborn
  • Hadoop
  • Jupyter
  • Google Colab
Web & databases
  • React
  • React Native
  • Node.js
  • Express
  • PostgreSQL
  • MySQL
  • MongoDB
Cloud & tools
  • AWS (S3, EC2)
  • Git
  • GitHub
  • Claude Code
  • Figma
  • VS Code
  • Notion

Recruiters & collaborators

Contact

Email

madhavarora00023@gmail.com