01 — Identity
02 — Projects
03 — Metrics
04 — Archive
05 — Skills
06 — Contact
FILE_001 // JAGTAP, R.
Classification · Digital Portfolio
Indexed · 2026
Adversarial
Rushda
Jagtap
InstitutionMITAOE, Alandi — Pune
Current CGPA
8.90
0 FIRST YEAR — CLIMBING 10
SpecialisationB.Tech CSE — Data Science
Validity
VALID FROM 2025
THRU
VALID THRU 2029
LinkedInin/rushda-jagtap
GitHubgithub/ruushhdaa

I build systems that
intercept threats
before they land.

B Tech CSE (DS) Sophomore Student. I work at the edge where data science meets cybersecurity. To me, they are not two separate fields—they are one single craft. I call it Cyber Adversarial Analytics. My focus is on building adaptive systems that map what "normal" looks like, allowing them to catch anomalies before a traditional rulebook ever could. I aim to engineer defenses that look for the patterns and ask the questions most standard dashboards miss. I prioritize real-world utility, think in patterns, and spend my time exploring the less-mapped territories of machine learning security. Outside of tech, I am driven by the same force to focus on patterns—whether I am reading, sketching, or out in nature.

// Live Status
DeployedNEITH — AI-powered NIDS with GNN + MITRE ATT&CK
Live atneith-green.vercel.app
TargetGSoC 2027 — OWASP / Honeynet Project
TargetIEEE ICCCE research publication
0% Detection accuracy
0 Transactions analyzed
0 Security systems built
0 CGPA — FIRST YEAR
↗ TRAJECTORY
01
NEITH
Network Entity Intelligence & Threat Hunter · AI-Powered NIDS
GraphSAGEGNNScapyMITRE ATT&CKNext.jsFlaskConformal Prediction
83.5% GNN Accuracy
28CICIDS Features
10sCapture Cycle
9ATT&CK Techniques

AI-powered Network Intrusion Detection System named after the Egyptian goddess who wove the fabric of reality. Captures live network traffic every 10 seconds via Scapy, computes 28 CICIDS-aligned features per IP, builds a real-time graph of devices and connections, and runs it through a trained GraphSAGE GNN to flag anomalous behavior. Detects attack patterns it has never seen before by learning what normal looks like.

Scapy Capture→ 28 Features/IP→ Graph Builder→ GraphSAGE GNN→ Anomaly Score→ MITRE Classifier→ Dashboard
⟐ Architecture Highlights
MITRE ATT&CK mapping across 9 techniques and 6 tactics. 90% Conformal Prediction confidence intervals — distribution-free uncertainty quantification. ADWIN drift detection monitors when network behavior shifts. SQLite persistence ensures alerts survive restarts. Reverse DNS + behavior-based role inference turns raw IPs into "github.com / external". Custom Egyptian war room dashboard — UnifrakturMaguntia, Cinzel, sacred geometry, zero rounded corners.
Live Mode
Real packet capture via Scapy with sudo privileges. 28 CICIDS features computed from raw traffic. Full GNN inference pipeline with StandardScaler alignment to training distribution.
Demo Mode
Generates realistic synthetic enterprise LAN data. Visitors experience NEITH working without root privileges or live capture. Deployed publicly on Vercel.
02
NIDS Deep Learning Engine
Network Intrusion Detection System · CT Lab Final Project
1D CNNTensorFlowKDD Cup 1999Multi-class
99.91% Accuracy
156,805Parameters
494KTraining records
5-classSoftmax output
Input(41×1)→ Conv1D(64)→ MaxPool→ Conv1D(128)→ MaxPool→ Flatten→ Dense(128)→ Dropout(0.3)→ Softmax(5)
⟐ Key Insight
Identified U2R precision drop (82%) as a dataset statistical artifact — only 12 test samples for that class — rather than a model failure. Perfect precision/recall achieved for DoS and Normal traffic.

Real-world network traffic classification engine built to catch zero-day and dynamic attacks. Compiled with Adam optimizer and sparse categorical crossentropy loss. Role: Lead Architect — designed, built, and evaluated the full pipeline.

03
SATARK
Smart Active Tracking And Recovery Key · Design Thinking Lab
FlaskSQLiteReactHoneypotQR

Cyber-physical identity verification system replacing physical documents with a single smart PVC card. Two operational modes — Active Defense (SAFE) and Passive Trap (LOST) — built on a Python Flask backend with dynamic state-changing QR codes.

SAFE Mode
Dynamic QR scan triggers push notification to owner — Allow/Deny. Verifier gets temporary 2-minute access window, then auto-expires and auto-deletes from their system.
LOST Mode
Stolen card scan shows fake "Verification Successful" to thief. Silently captures GPS coordinates, IP address, and device fingerprint — pings owner immediately.
⟐ Tagline
Satark Raho, Surakshit Raho. — Shortlisted for PCMC Police-sponsored project (Nigarani).
04
TrustSentinel
Live UPI Fraud Detection Engine · Datathon 2026
Random ForestSMOTESHAPStreamlitIEEE-CIS
98% Accuracy
590KReal transactions
118KUnseen test records
2-tierDetection system

Live UPI fraud detection utilizing individual behavioral profiling per card — not global averages. Two-tier system: hard rules (Tier 1) + Random Forest with SMOTE balancing (Tier 2). SHAP integration provides plain-English explanations of every flagged transaction.

⟐ Design Philosophy
Per-card behavioral baselines outperform population-level averages. A flagged transaction is meaningless without an interpretable reason — SHAP makes every decision explainable.
05
SHURI
LAN IoT Threat Detection · VELORA 1.0 — April 2026
NmapHoneypotsMITRE ATT&CKFlaskReact

LAN-deployed IoT threat detection system designed for low-compute hardware — runs locally on a laptop or Raspberry Pi with zero cloud dependency. Honeypots across Telnet, HTTP, MQTT, and SSH protocols. Custom "Blast Radius" scoring assesses network-wide impact rather than isolated per-device CVSS scores.

⟐ Framework Mapping
Every detected threat is mapped to CVEs, the OWASP IoT Top 10, and MITRE ATT&CK — providing context beyond a raw score. Flask API + React dashboard for real-time visibility.
06
Logic Gate Architecture
Digital Electronics Engineering · Hardware Lab Project
ANDORNOTNANDXOR

Hardware-level electronic gate system implementing all fundamental logic circuits. Physical breadboard construction using 7400-series ICs with truth table verification, K-map simplification, and Boolean algebra reduction.

7408 AND
7432 OR
7404 NOT
7400 NAND
7486 XOR
⟐ Why This Matters
Every neural network, every intrusion detection system runs on hardware that reduces to these gates. Understanding the physical layer isn't optional — it's the foundation. K-map optimisation taught me the same instinct I apply to model architecture: eliminate redundancy, minimise complexity.
07
This Portfolio
Custom Digital Portfolio · Pure HTML/CSS/JS
Vanilla JSCanvas APICSS GridZero Frameworks

Professional hub built with zero frameworks. Custom particle system via Canvas API. Scroll-reveal using Intersection Observer. Fluid responsive typography with CSS clamp(). Custom cursor, counter animations and typewriter effects.

❖

"Precision in logic.
Ruthlessness in defense."

❖
0%
Detection Accuracy
NIDS on KDD Cup 1999 · 10-epoch training
0
Transactions Analyzed
IEEE-CIS dataset · TrustSentinel
0
Systems Built
NEITH · NIDS · SATARK · TrustSentinel · SHURI · DEE · Portfolio
0
CGPA — FIRST YEAR
Sem I of VIII · Climbing
2/8 semesters completed
2026
NEITH — AI-Powered NIDS
GraphSAGE GNN · MITRE ATT&CK · Deployed on Vercel
Built and deployed a full AI-powered Network Intrusion Detection System with live packet capture, Graph Neural Network inference, MITRE ATT&CK classification, conformal prediction, and a custom Egyptian-themed dashboard.
Deployed · Live
Apr 2026
VELORA 1.0 Hackathon
Pitch: SHURI — LAN IoT Threat Detection
Presented SHURI, the LAN-deployed IoT threat detection engine with custom Blast Radius scoring, honeypots, and MITRE ATT&CK mapping.
Presented · View Cert ↗
2026
Nigarani — PCMC Police Project
Presented: All Projects.
Shortlisted for a PCMC Police-sponsored project.
Shortlisted
2026
Datathon 2026
Pitch: TrustSentinel
Pitched the TrustSentinel fraud detection engine.
Presented · View Cert ↗
2025–26
NIDS — CT Lab Final Project
Lead Architect · 99.91% Accuracy
Designed and built the full 1D CNN pipeline on KDD Cup 1999 dataset. Identified statistical artifacts in U2R class.
Faculty Presented
2025–26
Logic Gate Architecture — DEE Lab
Hardware Project · 7400-Series ICs
Built all fundamental logic gate circuits on physical breadboards. Verified truth tables, applied K-map simplification.
Lab Completed
// Certifications
IBM
AI Fundamentals
IBM
View ↗
NASS
COM
Digital Engineering 101
NASSCOM × MeitY
View ↗
CSCO
Python Essentials 1 & 2
Cisco
View ↗
ANTH
ROPIC
AI Fluency Suite
Anthropic
View ↗
CSCO
Data Analytics
Cisco Network Academy
View ↗
CSCO
Jr. Cybersecurity Analyst
Cisco Network Academy
View ↗
Certificate
ML / AI
TensorFlow · Keras
Scikit-learn
SHAP (Explainability)
SMOTE (Class Balancing)
Random Forest · 1D CNN
Isolation Forests
Graph Neural Networks (GNN)
GraphSAGE
Conformal Prediction
Scapy (Packet Capture)
Cybersecurity
Nmap · Network Scanning
Honeypot Deployment
MITRE ATT&CK Framework
OWASP IoT Top 10
CVE Mapping
Threat Intelligence
ADWIN Drift Detection
Reverse DNS Analysis
Languages
Python
C++
JavaScript
SQL
HTML · CSS
Tools & Frameworks
Flask · Streamlit
React
SQLite
Git · GitHub
Linux WSL
Google Colab
Docker
Next.js
Vercel

Intercept
me at:

// rushdajagtap07@gmail.com