Saif Elsaady | Portfolio

Projects

Featured Projects

Machine Learning Hardware Failure Classification System

Engineered an advanced dual CNN architecture for automated hardware failure detection using specialized models for continuous and discrete sensor data.

Implemented custom deep learning pipeline with 48→96→128 progressive feature extraction and strategic dropout regularization (0.1, 0.05 rates).
Developed a comprehensive transfer learning framework utilizing Adam optimization (5e-5 LR), class weight balancing for imbalanced datasets, and L2 regularization (0.0001).
Built production-ready MLOps pipeline with automated model versioning, timestamped backup systems, and real-time performance monitoring.
Created an interactive GUI framework enabling non-technical users to deploy ML models for industrial hardware diagnostics.

Submarine Mine Detection System - Machine Learning Approach

Achieved 88.89% accuracy in life-critical submarine mine detection using PCA-optimized Multi-Layer Perceptron neural networks.

Processed 60-dimensional sonar signals with only 3 missed mines out of 29 test cases for enhanced crew survival.
Engineered an optimal dimensionality reduction pipeline using Principal Component Analysis, reducing feature space from 60 to 8 components (87% reduction).
Implemented systematic hyperparameter optimization across 60 PCA configurations using scikit-learn MLPClassifier with 100 hidden neurons.
Delivered mission-critical defense technology solution with 10.3% false negative rate for hostile environment navigation.

Weather Station

Developed a mobile PCB with I2C/SPI sensors, power-efficient architecture, and field-tested calibration for environmental monitoring.

Built environmental sensor array using temperature, pressure, and wind modules.
Developed a power-saving design using switching regulators and rechargeable cells.
Designed the PCB in Cadence and passed DRC checks before fabrication.
Performed accuracy tests in varied conditions to verify sensor readings.

🛠 Hardware Experience

Flight Controller Development

Modeled IMU data using Kalman filtering, assisted with sensor fusion algorithms, and validated performance through real-world UAV testing.

Cleaned and visualized IMU data in MATLAB using Kalman filtering to support performance evaluation.
Researched aerospace specs and datasheets to inform design accuracy during subsystem development.
Assisted in integrating sensor fusion logic and PCB testing workflows for increased system reliability.
Managed project timeline with Gantt charts and Git version control to meet client milestones.

UAV Hardware Integration

Designed power and ESC connection systems using a custom PCB, configured Veronte Autopilot, and optimized layout for signal integrity.

Integrated Veronte Autopilot with GCS for stable flight communication.
Engineered power conversion from 22V to 9V and 5V for subsystem reliability.
Optimized system for weight balance and ESC thermal response.
Verified signals and outputs through bench testing and signal tracing.

Weather Station

Developed a mobile PCB with I2C/SPI sensors, power-efficient architecture, and field-tested calibration for environmental monitoring.

Built environmental sensor array using temperature, pressure, and wind modules.
Developed a power-saving design using switching regulators and rechargeable cells.
Designed the PCB in Cadence and passed DRC checks before fabrication.
Performed accuracy tests in varied conditions to verify sensor readings.

Adaptive Apparel PCB

Built a BLE-powered wearable on CY8CKIT-142 with low-power C++ firmware and user safety compliance.

Developed BLE firmware in C++ for safe and interactive user experience.
Integrated safety logic for use in children's wearable devices.
Designed and tested PCB layout for manufacturability and durability.
Optimized wireless signal strength and power draw for longer wearability.

Audio Recording Circuit

Created an op-amp based audio system using ISD1700, including playback, erase, and amplification logic.

Implemented audio processing using ISD1700 series chip.
Added analog preamp circuit using LM741 op-amps.
Benchmarked audio clarity and optimized op-amp gain settings.
Enabled playback, erase, and record cycling using digital logic control.

Servo-Controlled Testing System

Automated servo motion and sensor feedback with Arduino firmware and Serial Monitor controls.

Integrated servo motors with Arduino PWM for precision stroke control.
Added load cell and LVDT for real-time force and position tracking.
Enabled parameter input through Serial Monitor interface.
Programmed closed-loop system for automated test cycling.

Custom Temperature Sensor

Built a PCB + I2C sensor system with LED debugging and oscilloscope testing workflows.

Designed PCB with temperature sensor and pull-up resistors.
Added LED output for visual status monitoring.
Verified sensor accuracy using oscilloscope signal tracing.
Coded C++ firmware loop to read and validate temperature readings.

Peak Detector Network

Simulated diode-op-amp peak detectors for real-time signal measurement and testing instrumentation.

Designed op-amp circuit for capturing signal peaks.
Tuned component values for analog signal tracking.
Tested design across variable input frequencies using scope.
Applied in amplitude-modulated signal diagnostics.

Assistive Ball Launcher

Designed a switch-activated DC motorized assistive device with torque optimization and <$40 BOM.

Developed switch-based motor control interface.
Designed ergonomic frame using low-cost materials.
Optimized motor torque for consistent launching force.
Kept total bill of materials under $40 for accessibility.

Flyback Converter Power Supply

Designed and simulated a 5V, 30W isolated power supply using a high-frequency flyback topology and MATLAB-based modeling.

Implemented galvanic isolation and voltage conversion with a 40:1 transformer and switching architecture.
Simulated full converter performance in MATLAB, achieving <50mV ripple and 75.65% efficiency.
Optimized Schottky diodes, capacitors, and MOSFET selection to minimize conduction and switching losses.
Proposed GaN-based switching and advanced PWM control to increase future efficiency beyond 85%.

Noise Monitoring Subsystem

Designed a breakout board with op-amp-based analog filtering and embedded C++ firmware for dynamic audio monitoring.

Wrote C++ firmware to read analog signals and display real-time audio levels.
Designed op-amp circuitry with capacitor filtering to enhance signal clarity.
Used potentiometers to allow adjustable noise sensitivity in real-time.
Validated system performance using oscilloscope and ATE testing tools.

Electric Skateboard

Redesigned and optimized an electric skateboard through reverse engineering and performance benchmarking.

Performed speed, range, and braking tests to validate performance in real-world conditions.
Redesigned key components including motor mount, wheels, and battery placement for efficiency and safety.
Integrated headlights, carrying handle, and riding modes to improve usability.
Applied structured test plans to verify electrical and mechanical systems met design goals.

Automated Mini Garage Door

Built a servo-powered garage door prototype with bidirectional control and basic electrical circuitry.

Implemented servo motor system triggered by switch for door movement.
Designed and wired the circuit using breadboard, resistors, and jumpers.
Constructed wooden frame for structural support and safe enclosure.
Tested motion logic for consistent up/down operation and optimized power draw.

Digital Logic Gates

Simulated 6-bit logic gate operations in Verilog to validate binary arithmetic and hardware design.

Implemented AND, OR, XOR, NOR, and XNOR operations using register-based modules.
Used procedural stimulus generation in initial blocks to simulate inputs and outputs.
Monitored register changes with $monitor to track logic behavior in real time.
Validated functional correctness of each gate using waveform inspection.

Energy Conversion Launcher

Designed and built a rubber band-powered projectile launcher, optimizing elastic energy conversion for maximum efficiency.

Optimized energy transfer to maximize launch efficiency, reducing mechanical losses.
Validated theoretical vs. experimental results, refining design through real-world testing.

Fan Reverse Engineering

Redesigned and optimized a consumer stand fan for airflow, noise reduction, and assembly improvements.

Optimized propeller design to minimize noise by applying a thin plastic coating for dampening vibrations.
Redesigned fan cage for easy disassembly, improving maintenance and air quality by reducing dust buildup.
Enhanced airflow distribution by modifying the cage geometry, increasing cooling efficiency.

Microgrid & Power Flow Design

Used Xendee to simulate peak load, islanding, and voltage regulation in a hybrid solar/wind/diesel grid in Oahu.

Developed a GIS-based microgrid layout, modeling load profiles and renewable energy potential using real-world data.
Performed cost-benefit analysis using Lazard's levelized cost benchmarks for optimal long-term affordability.
Implemented peak shaving with battery storage and renewables to reduce dependency on the grid.
Conducted 3-day resilience analysis to ensure critical load coverage with backup generation.
Optimized dispatch scheduling to minimize operating costs while maintaining grid stability.

Power Flow Analysis | Power Systems Engineering

Simulated a resilient hybrid power grid for Oahu, Hawaii, integrating solar, wind, and diesel assets under both connected and islanded modes.

Created a one-line schematic in Xendee with feeders, buses, transformers, and microgrid inverters.
Performed peak load flow analysis to optimize transformer and cable sizing and correct voltage violations.
Simulated islanding scenarios to maintain voltage and reactive power balance using local generation.
Compared system behavior under grid-connected vs. islanded conditions with variable demand profiles.
Validated critical load coverage and power quality for uninterrupted operations during outages.

EV Charging Cybersecurity

Simulated attack vectors on onboard chargers using Arduino + MATLAB for both hardware and protocol testing.

Built hardware simulation to model EV charging protocol.
Wrote MATLAB scripts to simulate attack scenarios.
Emulated CAN bus signal interference and spoofing.
Verified vulnerabilities through test scenarios and logging.

💻 Software, AI & Simulation

UAV Traffic Management

Built Python-based flight optimization model for Honeywell Hackathon with 3D visualization and live path correction logic.

Developed real-time UAV path optimization using Python simulation.
Integrated 3D visualization for UAV airspace situational awareness.
Enhanced UAV safety via autonomous conflict resolution logic.

NVIDIA Stock & Volume Forecasting

Built time-series models (XGBoost, SVR, ANN) and applied PCA + t-SNE for forecasting optimization.

Applied supervised learning models for financial forecasting accuracy.
Performed EDA to visualize trends and trading behavior anomalies.
Used PCA and t-SNE to enhance input features and reduce dimensionality.

Instacart Basket Modeling

Applied ML and statistical regression to predict reorder behavior across departments and product types.

Performed regression and EDA on reorder patterns in Instacart datasets.
Built ML classifiers and conducted confidence interval testing.
Used probability modeling to optimize reorder prediction accuracy.

Customer Call Clustering (USAA)

Used NLP and K-Means to segment call transcripts, visualize friction, and improve service design.

Performed clustering and EDA on transcript data to segment call types.
Used n-gram and topic modeling to identify satisfaction drivers.
Presented insights in checkpoints to inform client-side UX changes.

Crime Prediction (ML)

Used clustering, logistic regression, and geospatial features to predict crime severity and classify weapon usage.

Trained classifiers to distinguish violent and armed crime categories.
Mapped geospatial trends with DBSCAN and choropleth clustering.
Modeled temporal patterns with Prophet for forecasting.

MATLAB Simulation

📐 Simulated Newton's Cooling Law, baton physics, and Lotka-Volterra equations for dynamics & biology models.

Built simulations of heat loss and motion in physical systems.
Modeled predator-prey behavior with Lotka-Volterra in MATLAB.
Visualized simulations with time plots and parameter sweeps.

Java Dice Game

Developed a simple object-oriented betting simulator using classes, input validation, and game logic.

Implemented OOP with custom Die and Main classes.
Validated input and tracked user balance during gameplay.
Used java.util.Random for dice roll simulation.

Adhesive Heat Testing

Conducted hypothesis testing on calorimetry data, comparing reported and experimental means.

Performed one-sample hypothesis tests using R.
Plotted confidence intervals and error bars for clarity.
Published findings showing statistical significance (p < 0.01).

EPL Goal Analysis

Applied normal distribution modeling to Premier League match data, testing fairness of reported stats.

Performed z-tests and statistical inference on EPL goal data.
Interpreted fairness and distribution validity of goal claims.
Visualized results with normal curves and deviation thresholds.

Energy Usage in AZ

Used regression to correlate energy burden to income tiers and analyze utility cost equity from RECS data.

Modeled relationships between household energy cost and income.
Visualized equity gaps using scatterplots and density maps.
Proposed targeted intervention strategies based on regression insights.

Retail Orders | ML & Statistical Project

Modeled grocery order behavior using regression, classification, and probability distribution.

Performed regression analysis to examine reorder and purchase trends.
Implemented ML models with scikit-learn for reorder prediction accuracy.
Visualized frequency distributions and PMFs using seaborn & matplotlib.
Applied probability modeling to evaluate checkout likelihood.
Proposed UX-driven data optimization strategies for cart behavior.

One-Sample Hypothesis Testing App

Developed Shiny R app for real-time hypothesis testing with CI visualization.

Built hypothesis testing functions with R (shiny, stats) for one-sample analysis.
Generated confidence intervals and statistical visuals dynamically.
Used ggplot2 to show hypothesis rejection with input-driven graphics.
Deployed fully functional app on Shinyapps.io for live demo use.

Abyssinian Style E-Commerce Website

Designed and launched Shopify site to support digital sales and enhance UX.

Implemented front-end with HTML/CSS and optimized mobile responsiveness.
Built product pages with integrated checkout and image optimization.
Customized navigation and branding for an intuitive shopping experience.
Worked with backend Shopify settings for inventory and SEO features.

UAV Camera Integration | Embedded Systems

Built live-streaming camera pipeline for UAV using Linux, FFmpeg, and OpenCV.

Developed video pipeline with OpenCV, GStreamer, and FFmpeg for low-latency feed.
Used Linux scripting for automation of encoding and secure data handling.
Streamed live video to GCS with reliability under UAV flight conditions.
Debugged and tuned performance for improved streaming quality.

Product Trends with R

Visualized product sales and inventory patterns using R data visualization libraries.

Created heatmaps and bar charts for category-level product distributions.
Used ggplot2 and tidyr for long-format transformations and mapping.
Highlighted sales performance drivers via aggregation and color maps.
Built interactive Google Slides for presentation and visual storytelling.

RF Signal Analysis | Statistical Inference

Explored RF detection trends in data center rooms using Python stats methods.

Cleaned and structured RF signal datasets using pandas for Python analysis.
Visualized signal behavior using matplotlib, seaborn, and distribution plots.
Calculated confidence intervals and performed inference on signal variance.
Applied chi-square and t-tests to estimate signal parameters.
Delivered reproducible results using statistical scripting workflows.

Machine Learning Hardware Failure Classification System

Engineered advanced dual CNN architecture for automated hardware failure detection using specialized models for continuous and discrete sensor data.

Implemented custom deep learning pipeline with 48→96→128 progressive feature extraction and strategic dropout regularization (0.1, 0.05 rates).
Developed comprehensive transfer learning framework utilizing Adam optimization (5e-5 LR), class weight balancing for imbalanced datasets, and L2 regularization (0.0001).
Built production-ready MLOps pipeline with automated model versioning, timestamped backup systems, and real-time performance monitoring.
Created interactive GUI framework enabling non-technical users to deploy ML models for industrial hardware diagnostics.

Advanced Signal Demodulation System - DSP/ML Engineering

Engineered comparative BPSK demodulation system using dual methodologies (classical DSP vs. neural networks).

Processed 10,000-bit transmission with SNR=1 Gaussian noise modeling for realistic communication channel simulation.
Implemented advanced classical demodulation pipeline using orthogonal basis projection onto cosine/sine vectors with trigonometric phase estimation algorithms.
Developed machine learning demodulation approach using optimized Multi-Layer Perceptron (10 hidden neurons) for direct signal-to-bit classification.
Bypassed traditional phase estimation for innovative neural network-based signal recovery.

Semiconductor Diode Modeling System

Engineered advanced semiconductor device simulation platform integrating fundamental physics equations with numerical optimization.

Achieved 1e-6 convergence tolerance across 30 iterative optimization cycles for precise parameter extraction from experimental I-V data.
Implemented robust Newton-Raphson circuit solver using SciPy optimization for nonlinear diode-resistor networks at 375K.
Developed multi-parameter optimization engine using iterative least-squares algorithms to extract barrier height, ideality factor, and series resistance.
Created comprehensive diode physics model incorporating temperature dependencies for accurate high-temperature semiconductor device characterization.

Submarine Mine Detection System - Machine Learning Approach

Achieved 88.89% accuracy in life-critical submarine mine detection using PCA-optimized Multi-Layer Perceptron neural networks.

Processed 60-dimensional sonar signals with only 3 missed mines out of 29 test cases for enhanced crew survival.
Engineered optimal dimensionality reduction pipeline using Principal Component Analysis, reducing feature space from 60 to 8 components (87% reduction).
Implemented systematic hyperparameter optimization across 60 PCA configurations using scikit-learn MLPClassifier with 100 hidden neurons.
Delivered mission-critical defense technology solution with 10.3% false negative rate for hostile environment navigation.

VLSI Circuit Optimization System, EDA Automation Approach

Engineered advanced HSPICE-driven circuit optimization platform automating parametric sweeps across fan factors and chain lengths.

Minimized propagation delay (t_phl) for 30pF high-capacitance loads through intelligent inverter sizing and topology optimization.
Developed sophisticated Python automation framework integrating industry-standard HSPICE simulation with dynamic netlist generation.
Enabled systematic exploration of 100+ circuit configurations with automated performance extraction and optimal design identification.
Implemented robust electronic design automation workflow using subprocess management for seamless HSPICE invocation and CSV parsing.

Heart Disease Prediction System - Machine Learning Approach

Achieved 94.44% diagnostic accuracy in heart disease prediction using Support Vector Machine with RBF kernel.

Outperformed 5 other ML algorithms on clinical dataset of 270 patient records with 13 cardiovascular features.
Conducted comprehensive exploratory data analysis using Pearson correlation matrices and covariance analysis.
Identified 6 critical predictive features with correlations above 0.41 for optimal model performance.
Implemented and optimized 6 machine learning algorithms achieving consistent 90%+ accuracy across multiple models.

AI Deployment Workflow Optimization

Led enterprise process transformation initiative reducing AI model deployment time by 75% through systematic IDEF0 modeling.

Reduced deployment time from 3-4 hours to <1 hour through automated workflow design for edge computing platforms.
Engineered comprehensive automation solution using CLI-based toolchain integration with $5,700 total solution investment.
Applied advanced enterprise modeling methodologies including Fishbone diagrams and 5 Whys analysis.
Designed standardized, version-controlled deployment pipeline for Jetson Nano and Coral TPU platforms.

EDA Compliance Validation GUI System

Engineered enterprise-grade desktop application featuring dynamic plugin architecture with importlib-based module loading.

Implemented sophisticated multi-check execution engine supporting variable schema CSV processing and intelligent data merging.
Developed advanced tkinter-based GUI framework with Excel-style multi-column filtering and real-time visualization.
Built production-ready software architecture implementing MVC, Observer, Factory, and Strategy design patterns.
Delivered scalable validation platform enabling non-technical users to execute complex compliance workflows.