Mohamed Saleh
PIB — Multilingual Press-Release Dissemination Pipeline
Selenium crawling → PDF-to-JSON → DistilBART summarization → translation into English + 9 Indian languages → TTS audio across 41 sample releases. Published at ICITIIT 2024, IEEE Xplore.
Projects
Distributed systems,
research & things I built.
Selected engineering projects, college research, hackathons and internships. Every project has a detail page.
Lung Cancer Detection via DNA Methylation
LASSO feature selection on 188 NCBI GEO samples (27,578 CpG sites) + ensembled RIPPER classifier via soft-voting — detects presence, stage and recurrence at 94.3% accuracy vs Random Forest 77.1%.
COVID-19 Helpline Chatbot
Keras Sequential intent-classification NN (7 intents, 32 training patterns, Nesterov SGD) with NLTK WordNet-lemmatized bag-of-words and a Tkinter desktop GUI.
Travel Document Management System
C++ OOP records & travel-eligibility system — Passport/VISA/Others inheritance hierarchy over a 14-field record, file-based persistence, 5 eligibility rules and 3 fee-based renewal/application flows.
Low-Bandwidth Multimedia Transfer System
Inverse-cosine compression + signal processing to transmit images/audio/video over 12.5 kHz / 9.6 kbps police radiosets with extreme bandwidth limits. 1st place in the internal hackathon.
Graphical Password Authentication System
Python/Tkinter visual credentials on a configurable grid — from a 389,112-combination 3×3 baseline — shoulder-surfing and brute-force resistant. Runners-up in the internal hackathon.
WearHelmet — Helmet-Detection & Challan System
YOLOv3 + CNN pipeline detecting helmetless riders from toll-plaza footage, with 4 evidence crops per violation, MongoDB evidence logging and accusation-number retrieval.
Analysisstant — Automated NSE Stock-Data Platform
Daily NSE bhavcopy ingestion of ~2,000 securities/day (~500K rows/year) into PostgreSQL with ISIN-based company enrichment and a Flask/Bootstrap analytics dashboard.