Reliable, efficient AI systems that can be evaluated rigorously.
I work at the intersection of LLM-based agentic systems, software engineering, and distributed machine learning. My current research focuses on guardrails, evidence-gated decision making, resource efficiency, and reproducible evaluation, alongside prior work in federated learning and communication-efficient edge AI.
Software engineering with systems thinking and applied AI.
I combine software engineering, system design, DevOps, testing, and Agile delivery with hands-on AI work. My experience includes complex and regulated environments, cross-functional collaboration, production-oriented problem solving, and translating emerging technology into practical technical solutions.
Current Research & Research Software
Current research and selected reproducibility artifacts. Formal publication status, venues, and manuscript details are listed separately under Publications.
Current and published work spanning reliable LLM agents, evidence-gated execution, resource-aware orchestration, federated learning, and communication-efficient AI. Where possible, research is accompanied by code or reproducibility artifacts.
Guardrails and Resource Efficiency in LLM-Based Agentic Workflows
Ongoing Master's Thesis Research · 2026Evaluating reliability and resource use in LLM agents.
See moreSee less
Investigating reliability and computational resource efficiency in LLM-based agentic workflows through a controlled comparison of conventional RAG, fixed-policy agentic workflows, and evidence- and resource-gated recovery. The study combines reproducible technical benchmarking with independent semantic evaluation and planned organizational research.
- Tools: LangGraph, LangChain, Retrieval-Augmented Generation (RAG).
- Methods: agentic workflow design, evidence-aware recovery, and abstention.
- Evaluation: LLM evaluation and benchmarking, resource and latency measurement, experimental design, and reproducibility.
R3Tune
Accepted Full Paper · IEEE FLLM 2026Evaluated in 4,000 runs across three open-weight model families. R3Tune met a pre-specified non-inferiority criterion for reliable correctness, without demonstrating meaningful token savings in the primary comparison.
Reproducibility artifactEvidence-Gated Budget-Aware Code Repair
Reproducibility Artifact · Submitted ManuscriptEvidence-gated code-repair workflow combining candidate generation, verification, and explicit resource constraints.
Reproducibility artifactuAMBIT
Reproducibility Artifact · Published ResearchReproducibility artifact for byte-budgeted update compression with error feedback in federated learning.
Code & reproducibilityAMBIT
IEEE ICFEC 2024Adaptive pruning research for efficient machine learning at the edge, forming the earlier foundation for subsequent work on communication- and resource-efficient distributed learning.
Budget-Aware LangGraph Agent
Research PrototypeResearch prototype exploring budget-aware planning and resource-constrained LLM agent execution.
View prototypeAcademic Service & Peer Review
Reviewer, UCC 2026
19th IEEE/ACM International Conference on Utility and Cloud Computing. Invited by a TPC member and accepted as a reviewer for UCC 2026. Completed and submitted one peer review evaluating technical contribution, architecture, methodology, experimental evidence, clarity, latency, and deployment trade-offs.
ConferenceExternal Reviewer, ECML PKDD 2026
Served as an External Reviewer for ECML PKDD 2026 and completed two peer reviews, assessing technical contribution, methodology, experimental evidence, clarity, and overall research quality.
ConferenceExpertise
LLM Systems & Agents
- Agentic workflow design
- Guardrails & evidence-gated execution
- Resource-aware LLM orchestration
- RAG and document-based knowledge workflows
- Reliability and failure-mode analysis
Experimental Evaluation
- Controlled baseline vs. intervention studies
- Accuracy, reliability, latency & resource metrics
- Benchmark design and evaluation rubrics
- Reproducible experimentation
- Quantitative analysis
Federated & Distributed ML
- Federated learning
- Distributed machine learning
- Communication-efficient learning
- Model compression & pruning
- Edge AI / resource-constrained ML
Research Engineering
- Python
- TensorFlow
- LangGraph / LangChain / LLM workflow prototyping
- Git & GitHub
- Research software & reproducibility artifacts
Software Engineering
- System design & software architecture
- Full software development lifecycle
- API integration
- Web application development
- Testing & quality assurance
DevOps, Cloud & Platforms
- DevOps practices
- Microsoft Azure
- Git & GitHub
- ServiceNow
- Data and platform integration
Applied AI
- LLM-based systems
- AI workflow prototyping
- Machine learning experimentation
- Model evaluation
- AI reliability & resource efficiency
Delivery & Collaboration
- Agile methodologies
- Scrum & Kanban
- Jira & Confluence
- Stakeholder communication
- Cross-functional collaboration
Security & Governance
- Information security awareness
- GDPR and privacy considerations
- Software & network security
- Risk-aware technical delivery
Technical Leadership
- Team facilitation
- Project planning
- Process improvement
- Technical communication
- Business / technology alignment
Academic Publications & Manuscripts
Peer-reviewed publications, submitted manuscripts, and manuscripts in preparation covering efficient distributed learning and resource-aware, reliable LLM workflows.
Manuscripts in Preparation
Guardrails and Resource Efficiency in LLM-Based Agentic Workflows: An Empirical Study of Document-Based Knowledge Work
An empirical study of LLM-based agentic workflows for document-grounded knowledge work, examining recovery policies, execution reliability, resource consumption, and evaluation consistency.
Accepted Full Papers
R3Tune: Stage-Aware Token Budgeting for Multi-Stage LLM Workflows
Reproducibility ArtifactSubmitted Manuscripts
Evidence-Guided Budget Allocation for Multi-Stage Large Language Model Orchestration
Reproducibility ArtifactPublished Research
uAMBIT: Byte-Budgeted Update Compression with Error-Feedback for Federated Learning
AMBIT: An Efficient Pruning Technique in Federated Learning for Edge Computing Systems
View PublicationMaster’s thesis (UiO, 2023)
Adaptive Mean Absolute Deviation Pruning in Federated Learning
View PublicationSelected Engineering & AI Projects
Selected hands-on work demonstrating applied AI, software engineering, workflow design, automation, and product-oriented technical problem solving.
Selected Academic Team Projects
Bachelor projects where I served as Scrum Master while also contributing to delivery and technical work.
MakerspaceLab — Bachelor Project (Software Engineering)
OsloMet · 2025
Served as Scrum Master for the 2025 Software Engineering bachelor project, supporting team coordination, iterative delivery, and the Scrum process.
Telia — Bachelor Project (Information Technology)
OsloMet · 2021
Served as Scrum Master in the Information Technology bachelor project, facilitating team delivery and contributing technically with Laravel and Onfido-related work.
Folcen — Mobile Communication Platform
Mobile Product Engineering · v1.0.0
Production-oriented Ionic/Angular mobile application combining real-time messaging, voice/video communication, native mobile integrations, and product-focused UX. Engineering work included UI refinement, guided interactions, alert behavior, release readiness, and Android delivery.
Selected product screens from the Folcen mobile application.
Reliable LLM Code Repair Pipeline
Research Engineering Prototype
Designed an LLM-based software-repair workflow that combines evidence collection, candidate generation, verification, and resource-aware execution rather than relying on unrestricted single-shot code generation.
- Evidence-gated decision flow before accepting code changes
- Automated candidate evaluation and verification
- Explicit cost and resource constraints
- Experiment-driven comparison of alternative workflow designs
Budget-Aware LangGraph Agent
AI Engineering Prototype
Built an agentic code-generation prototype using LangGraph, exploring how planning, tool execution, and explicit resource budgets can be combined in a controllable workflow.
- Graph-based agent orchestration
- Tool-driven execution workflow
- Budget-aware planning and resource constraints
- Structured multi-stage execution
Voice Insights AI — Speech and NLP Pipeline
Independent Engineering Project
Built a local audio-analysis tool using Whisper for transcription, DistilBART for summarisation, and DistilRoBERTa for text emotion scoring.
- Speech-to-text processing pipeline
- Summarisation with DistilBART
- Emotion classification with DistilRoBERTa
- Local Gradio interface
Femi Digital Safety Platform
Founder / Product Initiative
Initiated an early-stage digital platform concept focused on women's safety and community support, covering product direction, stakeholder dialogue, user experience, and prototype planning.
- Defined product concept and intended user journey
- Developed early web-based prototype material
- Explored stakeholder and public-sector collaboration
- Led early product and project direction
Professional Experience
Professional experience combining software engineering, production-oriented delivery, platform operations, Scrum leadership, and technical collaboration in complex environments.
Industry experience that informs my research perspective on reliability, governance, coordination, operational constraints, and responsible deployment of technical systems.
Combined hands-on software engineering, platform administration, production delivery, and Scrum leadership in a regulated banking environment. Worked across frontend development, contact-center technology, testing and deployment, operational support, and internal AI experimentation.
Software engineering and Scrum experience in a regulated banking environment, including production systems, operational constraints, AI prototyping, deployment processes, and cross-functional technical coordination.
- Frontend Engineering: Contributed to application development using React, implementing and maintaining user-facing functionality as part of the team's software-delivery work.
- Genesys Platform: Worked with Genesys-based solutions, including operational and administrative tasks such as user-access management, permissions, reporting, configuration support, and day-to-day platform activities.
- Deployment & Release: Supported and performed application deployments across S-test, A-test, and production environments, participating in the progression of changes from testing through production release.
- Quality Assurance: Designed and executed detailed test scenarios, including work related to Sbanken integration into DNB systems, helping validate changes before release.
- AI Initiatives: Developed and presented internal AI-oriented prototypes and technical initiatives, demonstrating potential use cases to colleagues and stakeholders through working demos. These initiatives were exploratory and were not adopted into production.
- Scrum Leadership: Led the team's daily stand-ups, sprint planning, sprint reviews, and retrospectives, including scheduling sessions, preparing agendas, facilitating discussions, and keeping ceremonies focused and productive.
- Team Coordination: Coordinated sprint activities, followed up actions and dependencies, supported removal of blockers, and helped maintain alignment throughout delivery cycles.
- Incident & Operational Support: Worked with incident resolution, troubleshooting, and service-management processes using ServiceNow in a production-oriented environment.
- Stakeholder Communication: Communicated technical progress, delivery status, risks, and proposed improvements to both technical and non-technical stakeholders.
Initiated an early-stage digital safety concept focused on women's safety and community support.
- Product Direction: Defined the intended user experience, platform goals, and early functional scope.
- Stakeholder Dialogue: Explored potential collaboration and communicated with the Norwegian Directorate for Children, Youth and Family Affairs (Bufdir).
- Prototype Planning: Translated the concept into an early digital-platform direction and demonstrable web material.
- Outcome: The initiative remained at concept/prototype stage and did not proceed to full product development.
Selected Education
Selected higher education spanning software engineering, systems architecture, artificial intelligence, and digital transformation.
Master in Digital Transformation
University of South-Eastern Norway (USN)
2024 - Present
Graduate studies in digital transformation, technology management, leadership, analytics, and responsible adoption of emerging technologies.
Master's thesis: Guardrails and Resource Efficiency in LLM-Based Agentic Workflows.
Ongoing research in 2026 on responsible LLM-based agentic workflows and resource efficiency.
Degree and thesis in progressMaster in Informatics: Programming and Systems Architecture
University of Oslo
Awarded: December 2023
Software engineering, systems architecture, distributed systems, and applied machine learning, including research in federated learning and model efficiency.
Master's thesis:
Adaptive Mean Absolute Deviation (AdMAD) Pruning in Federated Learning: A Hybrid Approach for Efficient Data Transmission and Model Performance
View thesis
Bachelor of Engineering in Software Engineering
OsloMet - Oslo Metropolitan University
Awarded: March 2026
Software engineering, web development, artificial intelligence, computer security, operating systems, and software project work.
CompletedAdditional completed education
Bachelor of Information Technology
OsloMet - Oslo Metropolitan University
Awarded: June 2021
Information technology, programming, web development, networking, databases, and applied AI.
CompletedShow additional education
Additional completed education
Bachelor in Informatics: Programming and Systems Architecture
University of Oslo
Awarded: June 2025
Programming, software engineering, databases, information security, and computer systems.
CompletedSelected Graduate Coursework & Research Projects
Weather Conditions and Monthly Personal-Injury Traffic Accidents in Oslo and Akershus
Spring 2026 · Open-data analytics study combining traffic-accident data from TRINE / Statens vegvesen with weather observations from the MET Norway Frost API. The analysis uses descriptive visualisation, Poisson and negative-binomial regression, model diagnostics, and a random-forest benchmark.
The Impact of AI-Based Tools on Students' Study Behavior and Perceived Academic Performance
Spring 2026 · Quantitative survey study examining relationships between students' AI use, perceived usefulness, study-behavior change, and perceived academic performance. The analysis includes descriptive statistics, reliability testing, Pearson correlations, and supplementary Spearman checks.
AI in Human Resource Management: Opportunities and Risks
2025 · Empirical course study examining the opportunities and risks of AI-supported HR processes, with emphasis on fairness, transparency, privacy, trust, and human oversight. The work combines relevant literature with questionnaire-based evidence and practical governance recommendations.
Agile Modernization in a Regulated Bank: Balancing Flexibility and Control in Legacy IT Transformation
Spring 2026 · Qualitative case study of agile and DevOps-oriented modernization in a regulated financial-services environment. The analysis examines legacy-system dependencies, governance, compliance, architectural guardrails, coordination, and the balance between team autonomy and enterprise control.
Articles
Writing on digital transformation, Agile delivery, project estimation, organizational design, and technology management.
Breaking Silos in Digital Transformation: My Experience
Published on LinkedIn, 2025
One of the biggest challenges in digital transformation isn’t technology—it’s alignment. Ambassadors bridge strategy and execution, a central office unifies efforts, and a compliance office keeps initiatives on-track and secure.
Read ArticleElevating Project Estimation with Three-Point Estimation & Monte Carlo Simulation
Published on LinkedIn, 2024
Combining Three-Point Estimation (Optimistic, Pessimistic, Most Likely) with Monte Carlo Simulation improves accuracy for complex, interdependent tasks—supporting realistic planning, risk management, and alignment.
Read ArticleExperience and Research-Based Insights on Agile Methodologies
Published on LinkedIn, 2024
Mapped Scrum, Kanban, XP, Lean Agile, Crystal, and FDD—then proposed effective hybrids (e.g., Scrum+XP, Kanban+XP) to balance structure, quality, and continuous delivery. Key theme: tailor frameworks to team and context.
Read ArticleApplying Ross et al.’s Five Building Blocks in Digital Transformation
Published on LinkedIn, 2024
Highlights shared customer insights, operational backbone, digital platform, accountability framework, and external developer platform—with banking examples that translate the theory into an actionable blueprint.
Read ArticleCertifications
Agile Project Management
Foundations of Project Management
Project Initiation: Starting a Successful Project
Applying Project Management in the Real World
Project Planning: Putting It All Together
Project Management Professional Certificate
PMI and NASBA
Managing Project Stakeholders
PMI and NASBA
Collaborating with Microsoft 365
PMI and NASBA
Managing Projects with Microsoft 365
PMI and NASBA
Optimizing Your Work with Microsoft 365
PMI and NASBA
Managing and Leading Developers
People Management
Coaching Skills for Leaders and Managers
People Management
Effective Technical Communication
People Management
Building Trust
People Management
Manager's Guide to Difficult Conversations
People Management
Inclusive Tech: Humane Code Reviews
People Management
Agile Software Development: Creating an Agile Culture
People Management
Succeeding as a First-Time Tech Manager
People Management
Moving from Developer to Engineering Manager
People Management
Volunteer Work
Connected the Nobel Peace Center initiative with students and university networks as a Student Ambassador for the Ignitor Community at the University of Oslo. Recognized for initiative, creative outreach, and going beyond the expectations of the ambassador role.
Certificate & RecommendationLed and supported the association's digital presence across multiple leadership terms. Designed and developed the organization's website, maintained technical resources, and supported digital communication and technical needs. The public website continues to credit the original website design to Mohammad Emad Hammami.
Supported refugees through information sharing, counselling, and community-based outreach as part of a voluntary refugee-support role.
Worked as an interpreter and supported information-sharing in settlement, health, education, social services, and other community-facing settings.
Delivered approximately 50 hours of social-studies instruction for adult learners, using Arabic-language support and facilitating classroom discussion.