Amr Alassal
Staff-level backend engineer for production AI systems, based in Amsterdam. Right now I'm building Booking.com's agentic customer-service AI: thousands of concurrent conversations, PII-guarded, sub-second.
Book a free intro15–20 min · No obligation · Online
- Python
- FastAPI
- LangGraph
- Kubernetes
- TypeScript / Node
- Kafka
- Postgres
- 12 years shipping production backend software
- 21 engineers across three teams, led hands-on at Bitvavo
- 2.6× cheaper LLM pipeline at 90%+ quality, Booking.com
- 95%+ automated quality gate on every AI summary, Booking.com
What I do
Most AI projects don't fail on the model. They fail on everything around it: the retry that never happens, the prompt nobody can change without a redeploy, the evaluation that only ever ran on the demo set, the monthly bill nobody predicted. My work sits in that gap, between a prototype that impressed the board and a system that survives Monday morning traffic.
I work in three areas, and most engagements touch two of them.
Agentic AI: multi-step LLM systems with reflection, parallel execution, guardrails, and quality scoring you can measure. The production workflow I built at Booking.com runs on LangGraph with reasoning models, with automated quality scoring at a 95%+ accuracy threshold.
System architecture: the services, queues, deployment and observability underneath, so the AI part is as reliable as the rest. FastAPI microservices, async batch processing, circuit breakers, Kubernetes with an Istio mesh across three regions.
AI strategy: what's worth building at all, what to buy instead, and what the first version realistically looks like. An executive MBA from Rotterdam School of Management means I can hold the business half of that conversation as well as the engineering half.
I stay hands-on. I write code in your codebase, and when an engagement ends your team owns something they can maintain without me.
Agentic AI
Multi-step LLM systems that hold up in production: self-reflection loops, async parallel execution, ML guardrails, automated quality scoring. At Booking.com that meant a ~11x latency reduction and >2.6x cost saving at 90%+ quality.
System architecture
The layer under the model: FastAPI microservices, async batch processing, circuit breakers, Kubernetes and Istio across three regions at 99.9% availability.
AI strategy
What to build, what to buy, and what it costs. An executive MBA on top of the engineering, so I can hold both halves of that conversation.
Track record
2025 – now
Booking.com · AI Staff Software Engineer (contractor)
- Production agentic workflow on LangGraph with reasoning models: iterative self-reflection, async parallel processing, automated quality scoring at a 95%+ accuracy threshold
- Agentic Context Engineering applied to the summarisation system: ~11x latency reduction and >2.6x cost saving at 90%+ quality
- Real-time PII masking pipeline with ML guardrails; Kubernetes across EU-NL/DE/UK at 99.9% availability
2024
Elevaide · Technical co-founder
- Secured €100K seed investment
- LLM meeting-analysis pipeline processing 100+ meeting hours
- React and Django product deployed on GCP with auto-scaling and fault tolerance
2022 – 2023
Bitvavo · Hands-on Engineering Manager
- Transaction system onboarding 57+ cryptocurrencies; Fireblocks custody securing €500M+ in ETH-based assets
- Scaled the engineering team 3x and cut incident rates 5x with a new incident framework
- Chainalysis integration for DNB compliance and AML tracking
2016 – 2022
Booking.com · Engineering Manager & Software Developer
- Cross-sell recommendations microservice serving the platform, with inventory across flights, cars, taxis and attractions
- Transport marketplace launch team; a CMS that cut transport launch time from days to minutes, scaled to 15 European cities
- Personalisation experiment on resort features worth >+10M in daily booking revenue; led the refugee internship programme
2015 – 2016
Amazon Web Services · Software Development Intern, Cape Town
- Personal health dashboard letting AWS customers monitor their deployed services, integrated with AWS internal monitoring
Who I work with
I work with three kinds of teams.
The first is a startup putting AI into the product for the first time. There's a demo that works, a founder who has promised something to customers, and no clear path between the two. That team needs someone who has shipped this before and can say which corners are safe to cut. I've been that team myself, as technical co-founder of Elevaide, where we built an LLM meeting-analysis pipeline on a €100K seed round.
Scale-ups come to me when the system has outgrown its first design. The AI feature works, but it's slow, or expensive, or nobody can change the prompts without breaking evaluation. Most of those are architecture problems. The summarisation system I optimised at Booking.com came out at ~11x lower latency and >2.6x lower cost at 90%+ quality, and that gain came from the context engineering and pipeline around the model.
The third is a larger organisation that has run its pilots and now wants an agentic strategy that survives contact with the roadmap, the budget and the compliance team. At Bitvavo I managed three engineering teams and delivered the Chainalysis integration for DNB compliance, so I know what a compliance team asks for.
Whichever of the three you are, the format is the same: I work in the codebase, next to your engineers. Book the intro call and tell me which one you are.
Amr Alassal
Staff AI backend engineer & consultant
About me
I grew up in Alexandria and studied computer engineering there. My first engineering job took me to Cape Town, where I spent six months at Amazon Web Services building a health dashboard that let customers see the state of their own deployed services. Then Amsterdam, and I stayed.
Since then I have taken on the systems other people find intimidating: the cross-sell recommendation service running platform-wide at Booking.com, three teams and 21 engineers at Bitvavo as a hands-on engineering manager, and agentic AI in production for customer service, where a wrong answer costs someone their evening.
The work I'm proudest of is the refugee internship programme I led at Booking.com in 2019, a four-month intensive front-end course for skilled people who had lost their footing through no fault of their own. Watching someone rebuild a career from that starting point permanently changed how I think about talent.
I did an executive MBA at Rotterdam School of Management while working full-time, because I kept ending up in rooms where the engineering was fine and the business case wasn't. Then I co-founded Elevaide, raised €100K, and learned what building without a safety net feels like.
Now I work independently, with a small number of clients at a time, and I do the work myself.
How an engagement works
Free intro call
15–20 minutes, online. You describe the problem; I tell you honestly whether I'm the right person for it and what I'd do first. No preparation needed, nothing to sign.
A scoped proposal
If it's a fit, you get a written scope within a few days: what I'd deliver, in what order, over what period, and what I'd need from your team. One page, no surprises.
We build
I work inside your codebase, next to your engineers. Weekly demos of something running, and a handover they can run without me: the code in your repo, the tests, and a written runbook.
Book a free intro
Pick a service and a time
Education
Executive MBA
Rotterdam School of Management, Erasmus University · 2021
Part-time executive programme completed alongside full-time engineering work. Managerial accounting, organisational behaviour, economics, business analysis, marketing, financial management, corporate strategy and operations.
Executive leadership communication
RSM, Erasmus University · 2024
Executive education in leadership communication.
Blockchain and Crypto applications
MIT · 2023
Applied blockchain and crypto-asset programme, taken during the Bitvavo years.
BSc Computer Engineering
Alexandria University · 2012
Four-year engineering degree (2007–2012).
FAQ
Are you available for freelance or consultancy work?
Do you work remotely or on site?
What language do you work in?
What happens on the free intro call?
How long is a typical engagement?
Can you sign an NDA, and who owns the code?
Contact
Use the form above: pick a slot for a free 15–20 minute intro call, or send an enquiry if what you have in mind is bigger than one conversation. Both land with me directly. I'm based in Amsterdam and work with teams across the Netherlands and Europe, online by default.
Location
Amsterdam, Netherlands