CYBERSECURITY / APPLIED AI / BUILDER

PIYUSH
SATYARTHI.

Exploring the intersection of cybersecurity, intelligent systems, and practical software. From AI research assistants and campus platforms to neural simulations, I build to learn and learn to build.

EXPLORE MY WORK ↗DOWNLOAD CV ↓GET IN TOUCH →
// Python · Security research · Local AI · Full-stack builds
OPEN TO COLLABORATIONSCROLL TO DECODE ↓   /   01 — 05
01 / WHO I AM

Curiosity is
my exploit.

A security-minded builder with an appetite for ambitious, practical software.

I work across offensive security, full-stack applications, and AI-powered tools — with a focus on useful systems that solve genuine problems.
PRIMARY FOCUSOffensive security & automation
CURRENT LEARNING PATHeJPT preparation · hands-on labs
WORK STYLEBuild → Test → Analyze → Improve
EDUCATIONBCA, Cybersecurity · Sharda University · 2025–28
FIELD NOTES / INDEPENDENT RESEARCHSELF-INITIATED · NOT A CLIENT ENGAGEMENT OR BUG BOUNTY PROGRAM
R.01

Broken object-level authorization

Observed BOLA / IDOR issues on multiple websites that exposed other customers' email addresses and order histories (OWASP API1:2023).

R.02

Business logic & authentication

Price tampering through missing server-side validation, authentication bypass to protected functionality, and account takeover via referral codes.

R.03

Recon, then manual validation

Subdomain enumeration → live hosts → endpoints and parameters, with automated output checked by hand in Burp Suite, ffuf, Nuclei, Subfinder and httpx.

R.04

Automation in Python & Bash

Scripts that automate reconnaissance and result collection. Affected sites and technical details are intentionally withheld.

02 / SELECTED BUILDS

Work worth showing.

A curated selection from my public GitHub. I read the repositories, READMEs, and implementation before writing these summaries.

REAL REPOSITORIES. REAL CODE. EXPLORE MY GITHUB ↗
00 — FLAGSHIP / CYBERSECURITY PLATFORM● LIVE / SELF-HOSTED
[ 00 ]

ZeroDay Care ↗

A self-hosted CTF and security-training platform where every student gets a private, disposable vulnerable lab to break into.

Built end to end: a Next.js + TypeScript frontend, a FastAPI + PostgreSQL + Redis backend, and a Docker lab service that starts a hardened container for each student, reachable only by its owner through an authenticating Nginx wildcard HTTPS proxy. 26 challenges cover SQL injection, XSS, CSRF, command injection, access control and crypto, including headless-browser “victim” labs, with per-student HMAC flags that can’t be shared.

Every lab runs non-root on a read-only filesystem with all capabilities dropped and no outbound network, inside a gVisor sandbox. Deployed on Microsoft Azure with Let’s Encrypt TLS, systemd services and tested daily backups.

01 / ISOLATED LABS02 / GVISOR SANDBOX03 / 26 CHALLENGES
NEXT.JSFASTAPIPOSTGRESQLDOCKERGVISORNGINXAZURE
VISIT THE LIVE PLATFORM SIGN-UP BY INVITE CODE
01 — FEATURED / AI SYSTEMS★ EDITOR'S PICK
[ 01 ]

FieldAssist ↗

An offline-first AI survival companion built to remain useful when the internet isn't.

Combines a local Ollama model with document retrieval powered by SentenceTransformers + ChromaDB. A FastAPI service answers questions from a curated knowledge base and requests routes from a local GraphHopper / OpenStreetMap server. Includes a React + TypeScript frontend.

01 / LOCAL LLM02 / RETRIEVAL03 / OFFLINE ROUTING
PYTHONFASTAPIOLLAMACHROMADBREACT + TS
VIEW SOURCE CODE READ THE README ↗
02 — FEATURED / RETRIEVAL ENGINEERINGOPEN SOURCE / PYTHON
[ 02 ]

Study Engine ↗

An AI-powered research companion that finds academic material and turns it into useful study insights through Telegram.

Built a Python Telegram bot and academic resource crawler with MongoDB Atlas hybrid retrieval, combining vector and keyword search. Uses Ollama / nomic-embed-text embeddings, document summarization and a configurable Ollama or Groq LLM backend. The repository also contains PDF, assignment and user-workflow components.

01 / RESOURCE CRAWLING02 / HYBRID SEARCH03 / LOCAL LLM
PYTHONOLLAMAMONGODBTELEGRAM BOTVECTOR SEARCH
VIEW SOURCE CODE EXPLORE SEARCH ENGINE ↗
03 — FEATURED / FULL-STACK SYSTEMSOPEN SOURCE / MULTI-ROLE
[ 03 ]

College ERP ↗

One connected academic workspace for attendance, grades, timetables and campus workflows.

Full-stack React 19 + Vite + TypeScript application with an Express/TypeScript + MySQL backend. Includes role-specific views for admins, HODs, faculty, students and parents; attendance analytics, marks and results, student timetables, hall tickets, and academic notifications.

01 / ROLE VIEWS02 / ATTENDANCE03 / MARKS & ANALYTICS
REACT 19TYPESCRIPTEXPRESSMYSQLZUSTAND
VIEW SOURCE CODE READ THE README ↗
04 — FEATURED / COMPUTATIONAL NEUROSCIENCEOPEN SOURCE / SIMULATION
[ 04 ]

FlyLab ↗

A fruit-fly brain simulation you can explore, stimulate and experimentally silence — directly in the browser.

A browser-based leaky integrate-and-fire simulation using the adult fruit-fly connectome, with 139,255 neurons and about 2.7 million connections. Sensory events feed into a virtual arena; users can disable neuron groups or regions to compare behavioral responses. Based on the FlyWire FAFB v783 wiring dataset.

01 / NEURAL DYNAMICS02 / LIVE STIMULI03 / LESION EXPERIMENTS
JAVASCRIPTSIMULATIONWEB WORKERSNEUROSCIENCE
VIEW SOURCE CODE HOW IT WORKS ↗
ALSO BUILT

More work from my CV. Not all of these are public repositories, so they are listed without links.

06 BUILDS
B.01AUTH / SECURITY

Web Authentication & Licensing System

Token-based licensing with hardware-ID bound authorization, admin token revocation and HWID reset. HMAC request signing, CORS restrictions and rate limiting; deployed behind Nginx and Cloudflare with UFW and TLS.

FLASKEXPRESSPRISMASQLITEMONGODB
B.02SMART INDIA HACKATHON 2026 · TEAM LEAD

QuantumMed

Led a six-member team building a hybrid quantum-classical heart-disease detection platform, with QSVM results framed against classical baselines.

QSVMQUANTUM MLHACKATHON
B.03

ASL Recognition

A convolutional neural network that classifies American Sign Language images, trained with TensorFlow and Keras.

TENSORFLOWKERASCNN
B.04

Instagram Chat Analyzer

A Streamlit app that parses exported Instagram conversations and turns them into charts.

PYTHONSTREAMLIT
B.05

IoT Prototypes

Arduino UNO R4 WiFi and ESP32 builds with gas (MQ-2), temperature and humidity (DHT22), motion (PIR) and ultrasonic (HC-SR04) sensors.

ARDUINOESP32
B.06

Heart Disease Prediction

A Random Forest model, written up as a coursework research paper in IEEE conference format, co-authored with N. Dhar.

RANDOM FORESTIEEE
MORE IN THE LAB

I also prototype authentication and security tooling. Projects 01–04 are selected from public repositories; ZeroDay Care runs live on my own server. Other experiments, security tooling and ongoing research live on GitHub.

ALL REPOSITORIES ↗
03 / MY TOOLKIT

The stack.

Move your cursor across the cards. Each technology gets a different motion inspired by its logo, shape, or purpose.

TECH // 001—020HOVER TO INTERACT
PythonBACKEND / AI01
JavaScriptFRONTEND / LOGIC02
TypeScriptTYPED UI03
ReactINTERFACES04
FastAPIAPI ENGINEERING05
SupabaseAUTH / DATABASE06
OllamaLOCAL MODELS07
DockerCONTAINERS08
Kali LinuxSECURITY LABS09
Burp SuiteWEB SECURITY10
GitVERSION CONTROL11
GraphQLAPI TESTING12
MongoDBVECTOR / DOCUMENT DATA13
ExpressREST API / BACKEND14
MySQLRELATIONAL DATA15
PostgreSQLRELATIONAL / PLATFORM DATA16
Next.jsFULL-STACK REACT17
NginxREVERSE PROXY / TLS18
RedisCACHE / RATE LIMITS19
AzureCLOUD DEPLOYMENT20
ALSO IN MY WORKFLOW
NMAPSUBFINDERNUCLEICHROMADBMONGODB ATLASNOMIC EMBEDDINGSSENTENCETRANSFORMERSGRAPHHOPPERLINUX CLIAMASSHTTPXFFUFGOBUSTERCURLFLASKPRISMASQLITESTREAMLITTENSORFLOW / KERASUFWSYSTEMDGVISORLET’S ENCRYPTARDUINO / ESP32
04 / WHAT'S NEXT

Always learning.

Progress isn't linear. Consistency beats shortcuts.

NOW / 2026

eJPT preparation

Working through networking, enumeration, exploitation foundations, and practical hands-on labs.

HANDS-ON PRACTICE

TryHackMe + tooling

Developing a systematic approach to Nmap, attack-surface discovery, and structured security testing.

ONGOING

AI-powered workflows

Building toward safe, reproducible security automation with clear human review and verification.

NEXT / DEEPER

Active Directory & privilege escalation

Kerberos and NTLM, Windows and Linux privilege escalation, and cloud security assessment: the next layer after eJPT.

05 / CONNECTION ESTABLISHED?

LET'S BUILD
SOMETHING.

Interested in cybersecurity, AI engineering, or collaborating on an ambitious product? Find the code on GitHub or connect with me.

GITHUB ↗LINKEDIN ↗