Hamza Halladj

Hamza Halladj

DevOps engineer and incoming Ph.D. student in Computer Science at Université Laval.

Now
DevOps engineer at Amelkis, running the company's move to Microsoft Azure
Next
Ph.D. in Computer Science at Université Laval, from January 2027
edge-cluster
CLOUD · KUBERNETES CONTROL PLANEAPI serverdesired stateCloudCoreKubeEdge, cloud sidecloud–edge tunnelEDGE · KUBEEDGE NODESedge-01ReadyNotReadyEdgeCorestreammqttedge-02ReadyNotReadyEdgeCoreinfermqttedge-03ReadyNotReadyEdgeCorestreaminfercamerasensorcameramapper
  1. t+0s edge-02 heartbeat missed
  2. t+2s edge-02 NotReady: node fault, not the app
  3. t+4s infer restarted on edge-03
What I'll research at Université Laval, sketched: notice when something at the edge fails, work out what failed, and recover the workload elsewhere.

~/about

About

I’m a DevOps engineer at Amelkis and an incoming Ph.D. student in Computer Science at Université Laval, where I start in January 2027 under the supervision of Prof. Mohamed Aymen Saied.

At Amelkis I migrated the company’s on-premise infrastructure to Microsoft Azure: about 150 resources defined as code in Terraform, plus the deployment pipelines, monitoring and security that keep them running. At the MISC Laboratory I run the lab’s own infrastructure too: Proxmox, OPNsense, LDAP single sign-on, Prometheus and Grafana.

My research is about keeping services up at the edge. At Laval I’ll work on availability management for IoT edge services on Kubernetes and KubeEdge: diagnosing faults and recovering across devices, workloads and nodes. Before that I worked on edge–cloud architectures and real-time violence detection, with one paper published and three under review.

I also teach: machine learning to 100+ students, and Master’s labs in distributed algorithms, operating systems and cloud computing.

~/work

Selected work

Systems I've built or run, drawn the way I'd sketch them on a whiteboard.

UsersHTTPSFront DoorWAF · TLS · public entryCI/CD pipelinesGitHub · Azure DevOpsVIRTUAL NETWORKAPPS SUBNETContainer Apps10 services + scheduled jobsvia private DNSDATA SUBNET · PRIVATE ENDPOINTSPostgreSQLAzure SQL MIRedisBastionadminVMs (rehosted)PLATFORMpush · OIDCRegistry (ACR)built imagesKey Vaultmanaged identityAzure OpenAImodel callsObservabilityLog AnalyticsApp InsightsGrafana · alertsAll of this is Terraform · ~150 resources · drift reconciled
General architecture, simplified: one public entry point, apps and data inside a private network, and code and secrets reaching the apps without passwords.

Amelkis · 2026

On-premise to Azure, as code

I moved the company's on-premise platform to Microsoft Azure, rehosting and replatforming it into about 150 resources defined in Terraform. The apps run in Container Apps inside a private network, the databases are reachable only through Private Endpoints, secrets live in Key Vault, and the pipelines sign in with OIDC instead of stored credentials.

Azure resources in Terraform
~150
services containerised
10
virtual users in k6 load tests
378

Azure · Terraform · Container Apps · Private Endpoints · Key Vault · OIDC · Azure DevOps · GitHub Actions · Managed Grafana

Cameraslive videoDOCKER · ONE CONTAINER PER SERVICEIngestionvideo framesAction modelbody · ResNetEmotion modelfacial emotionAccess controldoors · alertsframesframesscoreframesscorescoresApache Kafkaframesscores/metricsPrometheusscrapes every serviceGrafanadashboards
Frames go through Kafka to two models; their scores drive access control. Every service runs in Docker and exports metrics.

MISC Laboratory · ICSA 2025

Real-time violence detection

An event-driven pipeline that watches live video and flags violence from two signals at once: body movement, from ResNet spatial-temporal features, and facial emotion. Apache Kafka decouples video ingestion, the two inference engines and the access-control logic, and Prometheus and Grafana watch the whole stack.

detection accuracy
97%
signals: actions and facial expressions
2

Python · Apache Kafka · ResNet · Emotion recognition · Docker · Prometheus · Grafana

InternetRemote researcherWireGuardOPNsensefirewall · VLANs · DMZ · WireGuard VPNDMZVLANNginx reverse proxyautomated TLSWi-Fi Enterprise802.1XHTTPSPROXMOX VE · TERRAFORM + CLOUD-INITGit serverGPU CI runnersJupyterHubmulti-userLab servicesDocker · K8sFreeRADIUS802.1X authLDAPOpenLDAPsingle sign-onloginMonitoringPrometheusAlertmanager · GrafanaBlackbox probes
From the internet to a researcher's notebook: one firewall, one proxy, one directory, and monitoring on every hop.

MISC Laboratory · 2025 – now

A research lab's infrastructure

I run the lab's servers, network and identity. Proxmox VE hosts the virtual machines, provisioned with Terraform and cloud-init. OPNsense splits the network into VLANs and a DMZ and terminates a WireGuard VPN. OpenLDAP gives every service one login, an Nginx reverse proxy with automated TLS fronts Git with GPU CI runners and JupyterHub, and Prometheus, Alertmanager and Grafana watch all of it.

Wi-Fi users on 802.1X and FreeRADIUS
20–100
login for every service, via OpenLDAP
1

Proxmox VE · OPNsense · WireGuard · OpenLDAP · Terraform · Nginx · JupyterHub · Prometheus · Grafana

ClientgRPC · finds the leaderProduce · ConsumeGetServersNODE-1 · LEADERgRPC serverLogstore + indexseg 0seg 1activeRaftNODE-2 · FOLLOWERgRPC serverLogstore + indexseg 0seg 1activeRaftNODE-3 · FOLLOWERgRPC serverLogstore + indexseg 0seg 1activeRaftRaft: the leader replicates every writeSerf gossip · nodes discover each other and join the cluster
Three nodes: Serf for membership, Raft for replication, and segmented storage underneath.

Open source · Go

Dis-Log, a distributed commit log

A replicated, append-only log service in Go. Each node keeps records in segments, a store file plus an index of offsets. Raft replicates every write from the leader to its followers, Serf gossip lets nodes find each other and join the cluster, and clients speak gRPC, discovering the servers and the current leader through a custom resolver.

Go · gRPC · Protocol Buffers · Raft · Serf

  • Wi-Fi 6 802.11ax
  • Wi-Fi 5 802.11ac
  • Wi-Fi 4 802.11n
Mbit/s04080120160248163264128clients on one access point →Wi-Fi 6Wi-Fi 5Wi-Fi 4
Show as a table
ClientsWi-Fi 6Wi-Fi 5Wi-Fi 4
2139 Mbit/s119 Mbit/s91 Mbit/s
4154 Mbit/s128 Mbit/s99 Mbit/s
8137 Mbit/s117 Mbit/s90 Mbit/s
16139 Mbit/s116 Mbit/s85 Mbit/s
32137 Mbit/s113 Mbit/s90 Mbit/s
64131 Mbit/s117 Mbit/s83 Mbit/s
128133 Mbit/s116 Mbit/s90 Mbit/s
Mean TCP throughput between 1.4 s and 2.9 s of each run, read off the report's plots (about ±1 Mbit/s).

Network simulation · ns-3

Wi-Fi 6 vs Wi-Fi 5 vs Wi-Fi 4

I simulated 802.11ax, 802.11ac and 802.11n in ns-3: one access point on a 5 GHz, 40 MHz channel, with 2 to 128 clients sending TCP traffic. Wi-Fi 6 delivered the most throughput at every client count.

ns-3 · C++ · IEEE 802.11ax/ac/n · TCP

More projects

~/research

Research

Ph.D. project · Université Laval · starts Jan 2027

Availability Management for IoT Edge Services on KubeEdge

Keeping IoT services at the edge available on Kubernetes and KubeEdge: diagnosing faults and recovering across devices, workloads and nodes.

Supervised by Prof. Mohamed Aymen Saied

Publications

  1. 2025

    A Multimodal Approach for Real-time Violence Detection Based on Physical Actions and Facial Expressions

    H. Halladj, A. Bouzenada, D. E. Saidouni, M. R. Bennamoun, and A. Nouari

    Proceedings of the First International Conference on Smart Applications (ICSA 2025) · Springer · doi:10.1007/978-3-032-19929-4_10

    BibTeX
    @inproceedings{halladj2025violence,
      title     = {A Multimodal Approach for Real-time Violence Detection Based on Physical Actions and Facial Expressions},
      author    = {H. Halladj and A. Bouzenada and D. E. Saidouni and M. R. Bennamoun and A. Nouari},
      booktitle = {Proceedings of the First International Conference on Smart Applications (ICSA 2025)},
      publisher = {Springer},
      year      = {2025},
      doi       = {10.1007/978-3-032-19929-4_10}
    }

Under review

  1. 2026

    Title withheld during double-blind review

    H. Halladj, D. E. Saïdouni, G. Díaz, and B. Maati

    Journal · submitted to Computer Networks (Elsevier)

  2. 2026

    Title withheld during double-blind review

    H. Halladj and D. E. Saïdouni

    Conference · submitted to IAM 2026, 9th International Conference on Informatics and Applied Modelling (Guelma, Algeria)

  3. 2026

    Recency-Bounded Sequential Alerting for Edge Violence Detection

    H. Halladj and D. E. Saïdouni

    Conference · submitted to AICPS 2026 (ENS Oran, Algeria · IEEE)

Talks & workshops

  • Building Integrated IT Services for Research and Development

    ITRANS 2025 (Speaker) · University of Constantine 2

  • Introduction to Python Programming for Web Development

    WebExpo 2025 El Djazair (Workshop Instructor) · Constantine, Algeria

  • Smart Access Control via IoT: Contribution to the Smart City Project

    SENV'24 (Presenter) · MISC Laboratory

  • Create Your First Web Service Using Go

    GDG DevFest 2022 (Workshop Instructor) · Constantine, Algeria

~/experience

Experience

Professional

  1. Apr 2026 – Present

    DevOps Engineer

    Amelkis

    • Migrated the company’s on-premise infrastructure to Microsoft Azure (rehost and replatform), managed as code with Terraform: about 150 resources including VMs, SQL Managed Instance, PostgreSQL Flexible Server, Container Apps, Redis and Azure OpenAI
    • Set up network segmentation (VNet, subnets, NSGs, route tables), Private Endpoints with Private DNS, Bastion for admin access, secrets in Key Vault, and least-privilege RBAC with managed identities; reconciled configuration drift between Terraform and live resources
    • Manage the deployment pipelines: Azure DevOps, GitLab CI, and GitHub Actions using OIDC workload identity federation (no stored credentials) to build images to ACR and deploy to Container Apps
    • Built a multi-tenant dashboard for managing LLM token quotas per client
    • Observability: Log Analytics, Application Insights synthetic availability tests, metric alerts with action groups, Managed Grafana; automated TLS certificate renewal and rollout
    • Containerised the 10-service platform with Docker Compose and ran k6 load tests (up to 378 concurrent virtual users) for capacity planning

    Azure · Terraform · Container Apps · Azure DevOps · GitLab CI · GitHub Actions · Key Vault · Managed Grafana · k6 · Docker

  2. Dec 2025 – Present

    Federated Learning for IoT Intrusion Detection

    Research Project, MISC Laboratory

    • Designing a privacy-preserving simulation framework for IoT intrusion detection systems (IDS) using Federated Learning (FL)
    • Optimizing decentralized model training across heterogeneous IoT nodes to detect anomalies while minimizing data exposure

    Federated Learning · IoT Security · Intrusion Detection · Distributed AI

  3. Jan 2025 – Present

    Systems & Infrastructure Administrator

    MISC Laboratory

    • Run the lab’s virtualisation platform on Proxmox VE, with OPNsense for network segmentation (VLANs, DMZ), firewalling and a WireGuard remote-access VPN; VMs provisioned with Terraform and cloud-init
    • Centralised identity with OpenLDAP SSO; self-hosted Git with GPU CI runners, a multi-user JupyterHub for deep learning research, and an Nginx reverse proxy with automated TLS
    • Orchestrated containerized workloads using Docker and Kubernetes for centralized lab services
    • Monitoring and alerting with Prometheus, Alertmanager, Grafana and the Blackbox exporter (endpoint and certificate probes)

    Proxmox VE · OPNsense · Terraform · WireGuard · OpenLDAP · Docker · Kubernetes · JupyterHub · Prometheus · Grafana

  4. Dec 2024 – Jul 2025

    Real-time Violence Prediction & Access Control

    Research Project, MISC Laboratory

    • Architected a high-concurrency, event-driven microservices architecture using Apache Kafka as the message broker, utilizing asynchronous event streams to decouple video ingestion, inference engines, and access control logic
    • Engineered a multimodal deep learning pipeline integrating ResNet-based spatial-temporal features and facial emotion recognition, achieving 97% accuracy with optimized inference latency for edge deployment
    • Deployed a containerized processing stack using Docker, implementing full-stack observability with Prometheus and Grafana to monitor system health and ensure near-instantaneous detection

    Python · ResNet · Docker · Apache Kafka · Prometheus · Grafana · Microservices · Computer Vision

  5. Sep 2024 – Present

    Machine Learning Instructor

    Socode

    • Designed and taught a complete Artificial Intelligence program to 100–200 students
    • Covered supervised learning, regression, classification, neural networks, and Generative AI
    • Supervised hands-on Python projects, emphasizing practical implementation and ethical AI practices

    Teaching · Machine Learning · Python · AI

  6. Jun 2025 – Jul 2025

    Intern Supervisor — WiFi-Enterprise Infrastructure

    MISC Laboratory

    • Supervised interns deploying WiFi-Enterprise infrastructure for the lab
    • Supported 20–100 users using 802.1X, LDAP authentication, and FreeRADIUS

    802.1X · LDAP · FreeRADIUS · Networking

  7. Nov 2023 – Jul 2024

    Research Engineer — IoT Middleware & Access Control

    MISC Laboratory

    • Implemented formally verified, context-aware service discovery and collaboration protocols within a middleware for IoT-connected devices
    • Enabled self-healing and recovery in distributed environments
    • Integrated and deployed facial-recognition-based access control services as part of the Smart Lab project

    IoT · Middleware · Facial Recognition · Distributed Systems

Teaching

  1. 2024 – Present

    Lab Instructor

    Université de Constantine 2

    • Distributed Algorithms — consensus, fault-tolerance, and leader election mechanisms (Master’s level)
    • Advanced Operating Systems — process synchronization, concurrency, and inter-process communication
    • Cloud Computing — AWS Cloud Practitioner curriculum: cloud concepts, security, architecture, and billing

    Teaching · Distributed Systems · Operating Systems · AWS · Go

  2. 2024 – Present

    Teaching Assistant

    Université de Constantine 2

    • Assisted practical sessions for the Master’s course: Verification and Validation of Distributed Systems
    • Supported students applying formal verification, model checking, and testing methods to distributed architectures

    Formal Methods · Model Checking · LOTOS

Community

  1. Sep 2022 – Dec 2023

    Instructor & Community Organizer

    Google Developer Group (GDG) Constantine

    • Led workshops and events such as DevFest 2021–2022
    • Introduced participants to Go, web services, and cloud technologies
    • Promoted peer learning and collaboration through interactive, hands-on sessions

    Community · Go · Workshops

~/education

Education

  1. 2027 – 2030

    Ph.D. in Computer Science

    Université Laval, Québec, Canada

    Starts Jan 2027

    Thesis Availability Management for IoT Edge Services on KubeEdge

    Supervisor: Prof. Mohamed Aymen Saied

    Admitted for Winter 2027. Research on availability management for IoT edge services on Kubernetes and KubeEdge: fault diagnosis and recovery across devices, workloads and nodes.

  2. 2025 – 2026

    Ph.D. studies in Secure Cloud Architecture

    Université de Constantine 2 — Abdelhamid Mehri

    Thesis Development of Secure Integrated Systems for Smart Campus

    April 2025 – June 2026; not completed. Research focus: distributed cloud architectures, IoT systems, real-time AI, fault-tolerant distributed infrastructures, and intelligent access control.

  3. 2022 – 2024

    Master's in Networks and Distributed Systems

    Université de Constantine 2 — Abdelhamid Mehri

    Valedictorian — Ranked 1st university-wide

    Thesis Smart Access Control through IoT in MISC Laboratory

    Speaker at SENV'24 on smart access control via IoT; organized workshops (Create Your First Web Service Using Go) and major tech events (GDG DevFest 2022, Le Salon de l'Informatique 2023).

  4. 2019 – 2022

    B.Sc. in Computer Science

    Université de Constantine 2 — Abdelhamid Mehri

    Thesis Laptop Price Prediction Using Machine Learning for the Algerian Market

    Organized GDG DevFest 2021; active member of computer science student organizations.

~/skills

Skills

Cloud & IaC
Microsoft Azure, Terraform, AWS, cloud-init
Containers & CI/CD
Docker / Docker Compose, Kubernetes, Azure Container Apps / ACR, GitHub Actions, Azure DevOps, GitLab CI
Observability & Testing
Azure Monitor / Log Analytics, Application Insights, Prometheus / Alertmanager, Grafana, k6
Security & Identity
Azure Key Vault, RBAC / managed identities, OIDC workload identity, Private Endpoints / Private DNS, OpenLDAP SSO, TLS automation
Systems & Virtualization
Linux / systemd, Windows Server, Proxmox VE / VMware, Nginx, Git
Networking & Simulation
OPNsense / pfSense, VLANs / WireGuard VPN, FreeRADIUS / 802.1X / LDAP, IEEE 802.11ax/ac/n, ns-3 simulation, Network design & analysis
Languages & Frameworks
Python, Go, Bash / PowerShell, JavaScript / TypeScript, C / Haskell, Gin / Express.js / React, Ruby on Rails / Laravel
Distributed Systems & Data
Apache Kafka, Microservices / event-driven, Fault tolerance & consensus, IoT middleware, JADE agent platform, PostgreSQL / SQL Server, MongoDB / Redis
AI & Machine Learning
Scikit-learn / TensorFlow / PyTorch, Deep learning, Computer vision, Generative AI / Azure OpenAI
Languages
Arabic native, English IELTS 7.5 (C1), French TCF B2

~/contact

Contact

Email is the best way to reach me, whether it's about DevOps work, research or teaching.

I'm also on GitHub and LinkedIn.