AWS Solutions Architect

We are preparing to move a containerised application from a single-instance staging setup to a properly designed **production environment on AWS**, and we need an experienced architect to design it, size it, and tell us what it will cost. The distinguishing feature of this workload is a **CPU-bound biometric 1:N identification component**: every identity check compares a probe against the entire enrolled population, so compute cost scales with the number of enrolled users rather than with request volume alone. Getting the sizing and the cost curve right is the single most important outcome of this engagement. ### What you'll be working with (high level) - A handful of containerised services: an API, two independent matching services, a web front end, and a mobile client. - One relational database (MySQL). - Object storage for media, plus transactional email. - Today: deployed via CI to **EC2 instances running Docker Compose**, provisioned by hand, with **no infrastructure-as-code**. Full technical detail, source access and real data volumes are provided **after** an NDA is signed (see Requirements). ### Scope of work 1. **Production architecture design** — network topology, service placement, database strategy (managed vs self-hosted), secrets management, backup and disaster recovery, environment separation. 2. **Instance sizing** — concrete instance families and counts for each service, with the reasoning and the benchmark method behind them. We want to know *how you arrived at the numbers*, not just the numbers. 3. **Capacity & cost model for 1:N matching** — a model we can operate ourselves, expressing monthly cost as a function of **enrolled population size** and **checks per day**, at several population tiers. This is the core deliverable. 4. **Bottleneck and scaling analysis** — where this design stops scaling, what the mitigation options are, and roughly what each would cost to implement. 5. **Security & data protection review** — encryption in transit and at rest, key management, access control, audit logging, and data residency considerations. The data is sensitive personal information. 6. **Migration and IaC recommendation** — how to get from the current setup to the proposed one with minimal disruption, and what to codify first. ### Deliverables - A written architecture document with diagrams. - A cost model as a spreadsheet we can change the inputs on (not a static PDF number). - A prioritised, costed recommendations list. - A short live walkthrough session at the end. A precise deliverables specification is shared with the selected candidate before work begins. ### Requirements - Demonstrable experience designing **production** AWS environments, not just development ones. Please reference specific projects. - Real experience with **compute-bound, horizontally-scaled workloads** and with capacity planning you can defend with numbers. - Comfortable with EC2/ECS/EKS trade-offs, RDS, autoscaling, load balancing and AWS cost modelling. - Infrastructure-as-code (Terraform preferred). - Clear written English — the primary deliverable is a document a non-specialist stakeholder must be able to act on. **Nice to have:** prior exposure to biometric matching, AFIS, face recognition or any large-scale similarity-search workload. ### Mandatory before any access is granted Because this system processes sensitive personal data: 1. **A signed NDA.** We provide the document; it must be signed before any technical detail, repository access or infrastructure access is shared. 2. **Government-issued photo identification**, plus proof of address, for identity verification. Contractors who cannot or prefer not to provide these should not apply — no exceptions will be made. We will share the anonymised technical brief only after both are complete. ### Engagement - **Type:** fixed-price, milestone-based. - **Expected duration:** a few hours - **Start:** ASAP - Read-only access to a staging environment and to source code is provided for the assessment. No production access is granted at any point during this engagement. ### To apply, please answer 1. Describe a production AWS environment you designed where **compute cost scaled with dataset size** rather than request count. How did you model it? 2. What benchmarking approach would you use to size a service whose per-request CPU cost grows linearly with the number of stored records — before that dataset actually exists at scale? 3. When would you choose ECS over EKS, or plain EC2 with an autoscaling group, for a small number of long-running CPU-bound services? Give your actual reasoning, not a feature list. 4. How do you present a cost model to a non-technical decision maker who needs to choose between three growth scenarios? 5. Confirm explicitly that you are able to sign an NDA and provide government-issued identification. Applications that do not answer these questions will not be reviewed.