Engineering Intelligence Interface
AI, Telecom and Radiofrequency Engineer
Engineering intelligent systems across data, networks, software, and radiofrequency.
Python · Machine Learning · LTE/5G · RF Systems · Full-Stack Engineering
Connected engineering profile
Insaf’s work begins where disciplines stop being separate. Data, physical signals, network behavior, intelligence, and software are treated as parts of one engineering system.
Observe → interpret → decide → build
Master’s thesis · NTT DATA · Mar–Jul 2026
AI-Powered LTE/5G Network Monitoring and Incident Diagnosis Platform
NetSight AI transforms telecom network measurements into monitoring, alerts, diagnoses, and engineering recommendations. It combines real-time KPI analysis, telecom knowledge, anomaly detection, and root cause analysis in a modular full-stack platform.
Creates controlled telecom behavior for testing the platform end to end.
Creates realistic operating conditions before diagnosis begins. This turns the platform into a testable engineering system rather than a static interface.
Converts raw measurements into a structured operational view. Each KPI describes a different failure surface in the radio and transport chain.
A modular service layer receives, validates, stores, and exposes network information to the rest of the platform.
The diagnostic layer combines physical network knowledge with data-driven evidence. It does not merely label an anomaly; it builds a reasoned engineering interpretation.
The final layer turns engineering analysis into an operational workflow where network conditions can be inspected, explained, assigned, and reported.
System outcome Raw measurements become traceable engineering decisions.
Implementation evidence
Selected views from the implemented platform show KPI monitoring, cell analysis, root cause reasoning, incident management, and report generation.
Selected engineering case files
Each case begins with an engineering constraint and follows the reasoning path toward a testable implementation. Select a file to inspect its problem, approach, tools, and technical outcome.
RF device modelling + analytical extraction
A co-authored technical study connecting transistor physics, S-parameter extraction, model refinement, matching-network design, and nonlinear RF analysis.
Measured S-parameters describe terminal behaviour, but RF design requires a physically interpretable equivalent model that separates extrinsic parasitics from the active transistor core.
AI + RF modeling
Predict multidimensional RF behavior from geometry, material, and frequency-dependent inputs.
Electromagnetic simulations can be computationally expensive and generate large multidimensional datasets.
Train regression models on electromagnetic simulation outputs and use RF-informed features to estimate S-parameter responses across frequency.
A surrogate model can support rapid design-space exploration while keeping the original electromagnetic model as the physical reference.
Radiofrequency engineering
Design a stable amplifier with high gain and low noise near 10 GHz.
High-frequency gain is useful only when stability, impedance matching, and noise behavior remain controlled.
Perform transistor stability analysis, impedance matching, gain optimization, noise-figure evaluation, and controlled stabilization in Keysight ADS.
The project connects transistor-level behavior to system objectives, showing how gain, noise, and stability must be optimized together rather than independently.
AI + telecom planning
Estimate service quality while balancing coverage, demand, interference, and deployment constraints.
Base-station placement is a constrained systems problem, not a simple coverage-maximization task.
Build a regression-based model from planning variables to estimate service quality and support comparison between deployment configurations.
Machine learning acts as a planning support layer, helping compare complex alternatives without replacing the network-engineering constraints that define valid solutions.
Embedded signal systems
Build a reliable path from a physical sensor measurement to interpretable software analysis.
Sensor data is only useful when acquisition timing, communication, processing, and visualization form one reliable chain.
Connect sensor acquisition, embedded communication, signal processing, and software visualization into an end-to-end measurement system.
The project demonstrates full-chain thinking: every software result depends on the quality of the physical measurement and the reliability of the embedded link.
Biomedical signal intelligence
Extract meaningful indicators from physiological signals and convert them into real-time alert logic.
Physiological signals are variable, noisy, and time-dependent, so detection must be grounded in robust feature extraction.
Acquire physiological data, preprocess the signal, extract indicators, visualize trends, and trigger alerts when a defined pattern is detected.
The system shows how noisy physical signals can be transformed into interpretable events through careful preprocessing, feature design, and timing logic.
Environmental IoT
Continuously observe pH and turbidity while accounting for sensor drift and environmental variation.
Environmental measurements evolve slowly, drift over time, and can contain abrupt faults that must be distinguished from real change.
Acquire environmental indicators continuously, calibrate and clean the measurements, visualize temporal behavior, and detect anomalous deviations.
The platform connects field measurements to long-term data interpretation, where calibration and context matter as much as the detection algorithm.
Capability relationships
Select a system to trace the tools, engineering methods, and project evidence connected to it.
Builds models that remain connected to engineering context, measurable error, and explainable evidence.
Academic development
The academic path moves through electronics and RF toward telecommunications, software, and artificial intelligence, with each stage extending rather than replacing the previous one.
Faculty of Sciences, Abdelmalek Essaâdi University · Tetouan, Morocco
Faculty of Sciences, Abdelmalek Essaâdi University · Tetouan, Morocco
Morocco
Engineering methodology
Study the physical and technical behavior before selecting a solution.
Start with a clear and testable implementation that exposes the real constraints.
Measure performance, errors, limits, assumptions, and uncertainty.
Refine the system through testing, evidence, and controlled iteration.
Connection endpoint
Interested in intelligent networks, RF systems, machine learning, or engineering software? Let’s discuss the system you want to build.
Thank you for reaching out. Your connection request was transmitted successfully and will be reviewed as soon as possible.