PRANAV MAHESHWARI PRANAV0494

┌──(pranav㉿github)-[~]
└─$ whoami
Pranav Maheshwari — CS undergrad @ Graphic Era University, Dehradun · Class of 2027

┌──(pranav㉿github)-[~]
└─$ cat focus.txt
Machine Learning · Deep Learning (LSTM) · Time-Series Forecasting · Explainable AI (LIME & SHAP)
Training models by day, building full-stack apps by night.

`$ whoami --verbose`

🧠 Continuous authentication & behavioral biometrics — CADENCE: a transparent proxy that keeps verifying who is driving a session after login, fusing keystroke provenance, identity drift and automation signals through a sequential likelihood-ratio test
🛡️ Intrusion detection at scale — XGBoost & Random Forest across 6 public benchmarks (CICIDS2017, UNSW-NB15, CTU-13, BoT-IoT, IoT-23, BETH) with SMOTE and 5-fold cross-validation
📈 Time-series forecasting — ARIMA + LSTM for 30-day price prediction, with RSI/MACD indicators and Sharpe & VaR risk metrics
🩺 Patent filed — Offline-First Telemedicine Ecosystem for rural healthcare (RAG-based AI symptom detection + QR digital health records)
🏗️ Shipped a real-world school ERP during my frontend internship — React + Tailwind + Flask, role-based dashboards
🎤 Secretary @ GEU IEEE Student Branch — organized 50+ technical events & workshops
🎯 Open to ML / Software Engineering internships — let's talk!

`$ cat experience.log`

💼 Frontend Developer Intern — KV Public School, Meerut (3 months) Built a production ERP Management System: React + Tailwind frontends on Flask REST APIs, role-based dashboards, and cloud-backed data handling (SQLite / Google Sheets) automating attendance, fees, marksheets & transfer certificates.
🎓 B.Tech in Computer Science — Graphic Era (Deemed to be) University, Dehradun (2023 – 2027)
📜 Google Cybersecurity Professional Certificate — 8-course specialization: network security, threat detection & response, Linux, SQL, Python security automation
☁️ AWS training modules — Amazon HealthLake, Agentic AI (SNS, SQS, MQ), Generative AI sustainability

`$ ls skills/`

`$ cat projects.log`

🎼 CADENCE

Continuous Auth · Research Prototype

Every authentication control fires once, at the door. CADENCE keeps asking who is driving the session after login — a mitmproxy edge injects a capture SDK and reconciles form POSTs against observed keystrokes server-side, so patching the in-page JS yields a submission with no supporting keystrokes, which is itself the detection. Provenance, identity drift and automation fuse through a calibrated sequential log-LR (Wald's SPRT) into RFC 9470 step-up. 302 tests, and CI fails the build if any number in the docs disagrees with the measured-results file.

🛡️ Network Intrusion Detection Model Zoo

ML × Security

XGBoost & Random Forest classifiers trained across 6 public intrusion-detection benchmarks — CICIDS2017, UNSW-NB15, CTU-13, BoT-IoT, IoT-23 and BETH — using SMOTE for class imbalance and 5-fold cross-validation on multi-million-row captures, some GPU-trained via RAPIDS cuML.

🩺 Offline-First Telemedicine Ecosystem

Patent Filed

A healthcare platform for low-connectivity rural regions — RAG-based AI symptom detection, QR-based digital health records, and low-bandwidth teleconsultation, designed for scalable healthcare access where the internet barely reaches.

📈 Stock Forecasting & Risk

Time-Series · ARIMA + LSTM

30-day price forecasting combining a statistical ARIMA baseline with an LSTM sequence model, fed by RSI and MACD technical indicators and scored on Sharpe ratio and Value-at-Risk rather than accuracy alone — because a forecast that ignores downside risk is not a trading signal.

`$ htop --stats`

^{public + private activity · auto-updated daily by this workflow from the GitHub API}

📊 More stats — contribution activity graph

`$ ./snake --eat-contributions`

`$ ./connect.sh`

┌──(pranav㉿github)-[~]
└─$ echo "Thanks for stopping by — connection closed. 🟢"

Navigation Menu