M. A. DIRANY
Available for engineering roles

Mhammad Ali
Dirany

Computer & Communications Engineer
Embedded systems · Industrial automation · Applied machine learning

I build systems that run the whole way down: ladder logic and motor control at the bottom, ESP32 and Raspberry Pi firmware over MQTT in the middle, computer vision and React dashboards on top — then I write the report that explains all three.

Email me LinkedIn See the projects
AUST · Lebanon
CCNA certified
AR native · EN C2
Portrait of Mhammad Ali Dirany
MHAMMAD ALI DIRANYAGE 21
8ENGINEERING PROJECTS
2RESEARCH PAPERS
3+YEARS LEADING AN NGO
4COMMUNITY PROJECTS DELIVERED
About

Who I am

An engineer equally comfortable wiring a contactor, writing ladder logic, and training a neural network.

I'm Mhammad Ali Dirany, 21, a Computer and Communications Engineering student at the American University of Science and Technology (AUST) in Lebanon, academically advised by Dr. Gaby Abou Haidar.

My work sits where software meets hardware. For my senior design project I built a river waste collection and sorting machine end to end: a Delta PLC driving the conveyor, dual ESP32s and a Raspberry Pi 4 handling IoT and vision, a Kinect depth camera feeding a YOLOv8 classifier, and a React PWA over MQTT for monitoring. In parallel I ran a full DPCM/ADPCM compression study in PyTorch across speech and ECG data and wrote it up as an eight-page IEEEtran paper — keeping an honest negative result rather than a flattering one.

Outside engineering I co-founded NABA (جمعية نبا), a registered NGO in Qasarnaba, Bekaa Valley, where I serve as Vice President — leading proposal writing, donor outreach, and community programs in health awareness, child development, and economic empowerment.

I'm now focused on AI/ML engineering: agentic systems, retrieval-augmented generation, and applied computer vision, with an ICS/OT security auditing agent as my current flagship build.

  • Name Mhammad Ali Dirany
  • Field Computer & Communications Engineering
  • University AUST, Lebanon
  • Advisor Dr. Gaby Abou Haidar
  • Age 21
  • Email mhammad.dirany.eng@gmail.com
  • Phone +961 71 824 884
  • LinkedIn mhammad-dirany
  • Languages Arabic (native), English (C2 — ECPE Michigan)
  • Certified CCNA — Cisco Certified Network Associate
Skills

What I work with

Tooling I have actually shipped with — on benches, in labs, and in the field.

Domain

Industrial automation

PLC programming, ladder logic, motor control and safety interlocks.

  • Delta DVP-14SS2
  • WPLSoft
  • EASY 719-AC-RC
  • easySoft 6
  • Ladder logic
  • VFD / 3-phase
  • Modbus RTU RS-485
  • Stepper control
Domain

Embedded & IoT

Microcontroller firmware, sensors, telemetry and off-grid power.

  • ESP32
  • Raspberry Pi 4
  • MQTT
  • HiveMQ Cloud
  • libfreenect / Kinect
  • Solar & LiFePO4
  • MPPT
  • Firebase push
Domain

Machine learning

Applied ML — vision, sequence models and signal compression.

  • PyTorch
  • YOLOv8
  • LSTM
  • TFLite
  • MediaPipe
  • faster-whisper
  • DPCM / ADPCM
  • RAG · MCP · agents
Domain

Networking

CCNA-certified: routing, switching, addressing and troubleshooting.

  • CCNA
  • Routing & switching
  • VLANs
  • TCP/IP
  • Network design
Domain

Software development

Web, desktop and mobile front-ends over real hardware and real data.

  • React
  • PWA
  • Python
  • Swift / iOS
  • CustomTkinter
  • ffmpeg pipelines
  • Fusion 360 scripting
  • LaTeX
Domain

Leadership & communication

Getting projects funded, staffed and delivered — in two languages.

  • Proposal writing
  • Grant applications
  • Donor outreach
  • Stakeholder relations
  • Project coordination
  • Technical reporting
  • Bilingual AR/EN
Projects

Selected work

Eight projects across automation, machine learning, research and product. Open any card for the engineering detail.

RiverExtract

Senior design

AUST · Advised by Dr. Gaby Abou Haidar · Co-advised by Dr. Chafic Saide

A stationary, conveyor-based river waste collection and sorting system — built as a working prototype and specified as a full-scale, solar-powered installation.

Engineering detail
  • Prototype: 12V DC gear motor, Delta DVP-14SS2 PLC (WPLSoft ladder logic), dual ESP32s, Raspberry Pi 4, Xbox 360 Kinect (RGB/IR/depth) via libfreenect, solar-assisted 12V battery.
  • Full-scale design: 0.75 kW 3-phase AC motor with VFD, 3 m conveyor, automated colour-sorting bins with a stepper-driven sorting blade.
  • Power: 48V off-grid array — 3×550W panels, 60A MPPT, 50Ah LiFePO4.
  • Integration: Modbus RTU over RS-485 linking PLC to the IoT layer, MQTT telemetry, YOLOv8 waste classification.
  • App: React PWA monitoring spec — HiveMQ Cloud MQTT, Firebase push, E-Stop, jam recovery, bin-dump control, live AI detection feed and event log.
  • Deliverables: 60-page final report, A1 expo poster, single-page project site, Fusion 360 parametric Python script.
  • PLC
  • ESP32
  • YOLOv8
  • MQTT
  • React PWA
  • Solar

Tiny neural predictors in DPCM

IEEE paper

Cross-domain study — speech & ECG · 8-page IEEEtran

“Tiny Neural Network Predictors in DPCM: A Cross-Domain Study on Speech and ECG with Closed-Loop Stability and Evaluation Pitfalls” — a complete DPCM/ADPCM compression pipeline in Python and PyTorch.

Engineering detail
  • Five tiny neural predictor configurations (≤1500 parameters each) benchmarked against a well-tuned linear baseline.
  • Two domains: the Arabic Speech Commands Dataset and the MIT-BIH Arrhythmia (ECG) Database.
  • Closed-loop stability analysis, plus a catalogue of evaluation pitfalls that inflate reported gains.
  • Finding: no tested tiny network robustly beats the linear baseline under fair multi-slice evaluation — an honest negative result kept in the final paper.
  • PyTorch
  • DSP
  • DPCM/ADPCM
  • ECG
  • LaTeX

ICS-Sentinel

In progress

Agentic ICS/OT security auditor · AI/ML flagship

An autonomous security-auditing agent for industrial control systems, combining retrieval-augmented generation, Model Context Protocol tooling and a custom agent harness.

Engineering detail
  • RAG over ICS/OT vulnerability and standards corpora, so findings stay grounded in sources.
  • MCP tools give the agent real, auditable capabilities instead of free-form guessing.
  • Custom Hermes-Agent-style reasoning loop driving plan → probe → verify → report.
  • Scoped deliberately as a polished, demo-ready portfolio piece for AI/ML engineering roles.
  • RAG
  • MCP
  • Agents
  • ICS/OT security
  • Python

IoT water pump controller

Built

ESP32 + HiveMQ Cloud MQTT

A manual-start, auto-stop pump controller extending the RiverExtract IoT architecture — remote triggering with a stop condition that holds even when the network doesn't.

Engineering detail
  • Latch logic: PumpRun = (RemoteON_cmd OR PumpRun) AND NOT FloatFull.
  • Float switch guarantees auto-stop regardless of connectivity — safety in hardware, not in the cloud.
  • HiveMQ Cloud MQTT broker for remote command and status topics.
  • ESP32
  • MQTT
  • Control logic
  • IoT

Lebanese Student App

iOS · Swift

Native iOS app built with AI-assisted development

A fully native Swift app for Lebanese students — a community for sharing past exams and quizzes, plus a marketplace for selling skills and booking tutors.

Engineering detail
  • Built primarily with AI coding tools and minimal manual coding — an experiment in delivery speed.
  • Community module: upload, browse and search past exams and quizzes by course.
  • Marketplace module: students list skills or book tutoring sessions.
  • Swift
  • iOS
  • AI-assisted dev
  • Marketplace

Auto-clipping pipeline

Pipeline

Short-form video automation · Whop creator marketplace

A Python pipeline that turns long-form video into publish-ready vertical clips with no manual editing.

Engineering detail
  • faster-whisper for fast transcription with word-level timing.
  • Heuristic and LLM-based highlight scoring to pick the segments worth clipping.
  • ffmpeg for 9:16 cropping and burned-in captions, scripted end to end.
  • Python
  • faster-whisper
  • ffmpeg
  • LLM scoring

Sign language translator

Desktop / RPi

Word-level ASL recognition · Raspberry Pi 4 target

A Python desktop application that recognises word-level American Sign Language from a live camera feed, with an Apple-inspired dark interface.

Engineering detail
  • MediaPipe hand tracking for landmark extraction from the video stream.
  • LSTM sequence model exported to TFLite for on-device inference.
  • CustomTkinter GUI, designed for deployment on a Raspberry Pi 4.
  • MediaPipe
  • LSTM
  • TFLite
  • CustomTkinter

Visual positioning systems

In progress

Independent research · self-directed, no funding or lab access

A research paper on map-free visual relocalization — locating a camera without a pre-built 3D database.

Engineering detail
  • Survey and analysis of map-free / database-free approaches to visual relocalization.
  • Entirely self-directed: literature, experiments and writing without funding or lab resources.
  • Computer vision
  • Relocalization
  • Research
Background

Education & credentials

Academic track, certifications, and the research and leadership work running alongside them.

Degree · current

B.E. Computer & Communications Engineering — AUST, Lebanon

American University of Science and Technology. Senior design project RiverExtract, advised by Dr. Gaby Abou Haidar and co-advised by Dr. Chafic Saide.

Certification

CCNA — Cisco Certified Network Associate

Routing and switching, IP addressing, network services, security fundamentals and troubleshooting.

Certification

ECPE — Certificate of Proficiency in English (Michigan)

C2-level English. Arabic is my native language; I work and write technically in both.

Research

Author — IEEE-format paper on tiny neural predictors in DPCM

Cross-domain speech and ECG compression study in PyTorch, written up as an eight-page IEEEtran paper. A second paper, on visual positioning systems, is in progress.

3+ years

Co-Founder & Vice President — NABA NGO

Registered nonprofit in Qasarnaba, Bekaa Valley. Proposal writing, donor outreach and community program delivery across health, child development and economic empowerment.

Leadership

NABA — جمعية نبا

A registered Lebanese nonprofit in Qasarnaba, Bekaa Valley, running community development programs. I am a co-founder and its Vice President.

Funding & partnerships

  • Led donor and consortium outreach, and wrote the organisation's proposals.
  • Submitted an Expression of Interest to a consortium of SAFE, WeWorld, Mada and the Lebanese Observatory for Workers and Employees Rights — proposing a project combining GBV and child-protection awareness, financial-management training, and rose-soap production skills training for women and local farmers.
  • Prepared and submitted a MoSA (Ministry of Social Affairs) grant application centred on a Damascus rose value-chain project.

Programs & track record

  • Oversees NABA's annual Summer Camp — child intellectual development and mental health.
  • Programs across health awareness, child development and economic empowerment.
  • Cited in funding applications: 3+ years operating, 4 completed projects, average project value ≈ $5,000 USD.
Contact

Let's build something

Open to AI/ML engineering, embedded systems and automation roles, as well as research collaboration. Email reaches me fastest.

mhammad.dirany.eng@gmail.com +961 71 824 884 LinkedIn