divyansh@void:~$ cat /etc/operator
NAME ......... Divyansh Gupta
ROLE ......... Security Engineer · Full-Stack & AI Systems Builder · Blockchain Auditor
BASE ......... CS @ VIT — class of 2027
RECORD ....... AI-HR backend @ XtraGrad · Smart-contract auditing @ IBM · Frontend @ 1Stop.ai
STACK ........ Python · FastAPI · Next.js · CrewAI · multi-agent LLM pipelines · WebGL
HONOURS ...... Hack Energy 2.0 Finalist · PSB Hackathon 2026 (UCO Bank × IIT KGP)
DOCTRINE ..... assume hostile input · least privilege · leave an audit trail
MISSION ...... build the systems, then break them before anyone else can- 🔭 I build end-to-end platforms that assume hostile input — validated UIs, modular Python/FastAPI backends, and offensive tooling that maps attack surface
- 🤖 I architect multi-agent AI systems — CrewAI pipelines (TrustSphere AI — 4-agent deepfake forensics), LLM integration (Groq · Claude · Llama 3.3), spaCy / sentence-transformer semantic matching, async inference queues
- ⛓️ Audited Ethereum smart contracts at IBM — reentrancy, access control, gas-misuse
- 🕸️ I understand how apps are built and how they break
| module | payload |
|---|---|
| 🤖 AI systems | Multi-agent pipelines (CrewAI) · LLM integration (Groq · Claude · Llama 3.3) · spaCy · sentence-transformers · HuggingFace · semantic matching · Grad-CAM XAI |
| 🧠 Backend & data | FastAPI · async SQLAlchemy · Celery + Redis inference queues · risk-signal classification · web crawling · ElasticSearch |
| 🎯 Offensive security | Web-app pentesting · attack-surface mapping · recon & exploit automation · API/parameter fuzzing |
| 🛡️ Defensive doctrine | OWASP Top 10 · threat modeling · strict input validation · secure-by-design reviews |
| ⛓️ Blockchain security | Solidity auditing · reentrancy / access-control / gas-misuse detection · AST static analysis |
🟢 CASE FILE 01 — CivicShield · AI threat-intelligence platform (Hack Energy 2.0 Finalist)
Mission brief: real-time global threat dashboard — live DDoS/ransomware/phishing feeds on a 3D globe, auto-classified severity, vulnerability scanner, attack-surface mapper, and AI-driven scan intelligence.
Stack:
PythonJavaScriptAI/MLREST APIs3D VisualisationDeploy: 🌐 live · 📁 source
⚡ ACTIVE OP — TrustSphere AI · multi-agent forensic deepfake detection (in development)
Mission brief: four parallel CrewAI agents (Image · Video · Audio · Metadata) that detect deepfakes and classify the manipulation type — face swap, voice clone, GAN-generated. GAN fingerprinting attributes synthetic media to specific models (Stable Diffusion · DALL·E · Midjourney), with Grad-CAM XAI heatmaps, a Polygon blockchain audit trail, and a Chrome extension for right-click verification.
Stack:
FastAPICrewAIHuggingFaceNext.jsPolygonOpenCVStatus:🚧 in active development— multi-agent orchestration + XAI explainability layer
🟣 CASE FILE 02 — CipherMind · end-to-end encrypted AI chat
Mission brief: every message is AES-256-GCM ciphered before leaving the browser — zero plaintext in transit. PBKDF2 session keys, HMAC-SHA512 integrity, Groq Llama 3.3 70B backend with the API key fully server-side.
Stack:
ReactTypeScriptAES-256-GCMPBKDF2HMAC-SHA512Groq APIDeploy: 🌐 live · 📁 source
🟠 CASE FILE 03 — ChainAudit · automated Solidity security auditor
Mission brief: static-analysis engine that parses Solidity, extracts the AST, and hunts reentrancy, overflow, unchecked calls, broken access control, and gas-griefing — then ships structured audit reports with Critical→Low severity and remediation guidance.
Stack:
SolidityJavaScriptAST AnalysisWeb3EthereumDeploy: 🌐 live · 📁 source
🔴 CASE FILE 04 — SecureScout · enterprise security scanner (PSB Hackathon 2026 — UCO Bank × IIT KGP)
Mission brief: production-grade scanner for banking systems — vulnerable dependencies, hardcoded secrets, weak crypto, injection flaws. Next.js dashboard with 5-tier RBAC, immutable audit logging, and a CI/CD gate that fails the pipeline on Critical findings.
Stack:
PythonTypeScriptNext.jsExpressPostgreSQLRedisDeploy: 🌐 live · 📁 source
🕷️ CASE FILE 05 — Attack Surface Analyzer · offensive web recon engine
Mission brief: crawls targets, extracts endpoints & parameters, flags SQLi / XSS / IDOR candidates, and generates professional pentest reports — recon to report, fully automated.
Stack:
PythonRequestsBeautifulSoupRegexAutomationDeploy: 📁 source
🟡 CASE FILE 06 — Accident Risk Model · ML severity & blackspot prediction
Mission brief: end-to-end ML pipeline predicting 2026 road-accident severity and geographic blackspots — EDA, multi-factor feature engineering (time, weather, road, vehicle), and interpretable risk-tier scoring.
Stack:
Pythonscikit-learnpandasEDAFeature EngineeringDeploy: 📁 source
Found a vulnerability in my profile? Report it over coffee. ☕
"To break an application, you must first know how to build it."
