Hiring and team growth
Own hiring loops, technical interviews, onboarding, and mentoring so new and existing teammates build product context quickly and contribute across the engineering stack.
Tech Lead · London, UK
Dhia Kennouche
I lead an engineering team from architecture through delivery across card issuing, Open Banking, real-time transaction pipelines, and ML/GenAI solutions. I bring over a decade of hands-on experience across academic research, backend, web, mobile, cloud, and product delivery.
Leadership
Examples from the past year-plus as Tech Lead at Expend, where I have worked for five-plus years.
Technical direction
Shape roadmap planning, cross-team technical decisions, delivery sequencing, and on-call ownership while remaining hands-on in the codebase.
Engineering quality
Strengthen delivery confidence through pragmatic standards for testing, maintainability, observability, and technical decision-making.
Delivery ownership
Coordinate product, engineering, operational, and customer-facing work so complex changes move from ambiguous problem space to production safely.
Selected work
Projects I built or led over the past five-plus years at Expend.
AI-assisted OCR
In productionEvolved the OCR pipeline into a more AI-driven financial-document capture flow by migrating to Mistral OCR and then moving it onto Amazon Bedrock, improving how Expend extracts structured data from customer documents.
Read the detailsHide details
- Migrated document extraction to Mistral OCR, then switched it to run on Amazon Bedrock.
- Shipped the work into production as part of the AI Automation product surface.
- Focused the pipeline on higher-quality extraction for receipts, invoices, and other financial documents.
- Reduced Document OCR spend by 23x while increasing measured accuracy on edge-case documents to ~95%.
Supplier matching
Designed the algorithm, architecture, and backend that resolve clean merchant names from transactions and financial documents, adopting TypeSafe's Jev decision models in production shortly after release.
Read the detailsHide details
- Achieved near-100% accuracy for merchants the system has seen before.
- Made adding correct supplier names simple, with every customer benefiting from a new supplier instantly.
- Carried the feature across platforms: wrote the mobile code and guided the web implementation.
Customer insights AI assistant
Alpha testingDesigned and built solo: a LangGraph agent that answers customers' questions about their own data in natural language.
Read the detailsHide details
- Translated questions into SQL using schema-aware context and query validation.
- Added two guardrail layers: an LLM-as-a-judge check and deterministic SQL-policy validation.
- Built an event-driven question-answering engine so customer questions can be answered asynchronously, with conversations continuing across web and mobile.
Card issuer rollout and customer migration
Co-owned Expend's move to a new card issuer, taking joint responsibility for vendor and architecture decisions while owning the customer-facing rollout and migration work.
Read the detailsHide details
- Partnered with the CTO on vendor research, technical due diligence, and selection against product, risk, and operational requirements.
- Drove architecture alignment across the team for an event-driven system built to process millions of transaction events efficiently.
- Owned the frontend transition so customer-facing changes were clear, timely, and minimally disruptive.
- Owned migration execution for thousands of cards, coordinating new card creation, dispatch, customer communications, funds movement, and operational cutover.
Expense approval flows
Built the approval state engine behind Expend's configurable expense-review workflows.
Read the detailsHide details
- Modelled complex approval policies that assign the right reviewers based on requester, team, amount, expense type, policy breach status, tracking groups, and category rules.
- Handled ordered review steps, auto-approval, fallback reviewers, all-reviewer approval, and review-state visibility across the inbox and expense history.
- Designed and built the engine to process millions of expenses and transactions efficiently while scaling up and down automatically.
Open Banking reimbursements
Led the Open Banking integration behind reimbursing approved expense and mileage claims from a company bank account.
Read the detailsHide details
- Owned architecture, delivery, and maintenance across backend and web.
- Designed consent and authorisation flows, resilient synchronisation, and bank-data normalisation.
Card Connect
Built the external-provider integration for Card Connect, bringing real-time transaction data from connected business cards into Expend.
Read the detailsHide details
- Designed event-driven ingestion for card-transaction events as they happen.
- Built for the ordering, idempotency, and reconciliation guarantees financial data demands.
React-to-Next.js migration architecture
Designed the deployment and routing architecture for a phased migration from a legacy React app to Next.js.
Read the detailsHide details
- Enabled the legacy React app and the Next.js app to be deployed and served side by side.
- Created a navigation approach that let users move between both applications cleanly while migration work continued incrementally.
Focus areas
Card issuingExpense managementApproval workflowsFinancial transactionsOpen BankingReal-time payments dataEvent-driven architectureFrontend architectureLLM agentsLLM fine-tuningAWSTeam leadership
Research & publications
Peer-reviewed work on measuring and estimating software size with language models, and on how microservice teams actually work.
Automating software size measurement with language models: Insights from industrial case studies
H. Ünlü, S. Tenekeci, D. E. Kennouche, O. Demirörs · Journal of Systems and Software, vol. 231, 2026
Microservice-based projects in agile world: A structured interview
H. Ünlü, D. E. Kennouche, G. Kilinç Soylu, O. Demirörs · Information and Software Technology, vol. 165, 2024
My MSc thesis explored fine-tuning and evaluating language models to estimate software size from requirements written in natural language. The full publication list is available on DBLP.
Earlier work
I started with Ruby on Rails and Python as a teenager, contributing to open source and placing as a finalist in Google Code-in. My first Python contribution dates to 2011. Since then I have worked across Android, React Native, monolithic and microservice Java and Python systems, learning the trade-offs that inform how I lead today.
Core technologies
