Case study

PoseLight. AI headshots with human review.

PoseLight creates professional headshots for whole teams from phone selfies. We built the cloud platform, member apps and reviewer tools around that journey, with quality checks before generation and a human decision before delivery.

Practice AWS Cloud · DevOps Engagement Platform build · image generation · mobile apps
Delivery target
24 hours
At least 95% of orders within 24 hours of the final selfie upload.
Gallery target
50+
Frames per person across five studio styles.
Designed capacity
100K+
Generated images per month in the sizing model; not observed traffic.
Proof of concept
6 weeks
Likeness and quality evaluated with reviewers and a pilot team.
The brief

The promise depended on likeness, deadlines and privacy.

Whole-team orders arrive in waves. PoseLight needed to turn varied phone selfies into consistent headshots while giving reviewers enough time to judge each batch before the delivery deadline.

Poor inputs could waste generation time and leave members waiting for a re-upload request. Face images also needed controlled access, clear retention rules and an operational process that matched the privacy promise.

What we did

Put quality gates around the generation step.

Check the selfie while the member can fix it

Guided mobile capture checks framing, lighting and blur. Upload processing removes location metadata, scans files and uses a vision-based quality gate before the order proceeds.

Queue the work around team demand

Event-driven orchestration separates uploads, generation and scoring. Managed AI on Amazon Bedrock and ECS Fargate workers support bursts without a permanently running GPU fleet.

Give reviewers a ranked set of candidates

Automated quality and likeness checks rank the generated frames. A dedicated console lets PoseLight's reviewers approve, reject or regenerate them, with queues ordered by delivery deadline.

Make retention an operating control

Private galleries use signed delivery URLs. Lifecycle rules are configured to delete selfies after 30 days and galleries after 90 days, with audit records supporting retention reviews.

The outcome

One workflow from team invitation to reviewed gallery.

The engagement delivered member apps, a manager portal, a reviewer console and the generation pipeline. A pilot team and PoseLight's reviewers evaluated likeness and quality during the proof of concept before the production build-out.

The delivery objective is at least 95% of orders within 24 hours of the final upload, with 50 or more frames per person. These are service targets; the capacity model of more than 100,000 generated images a month is not a measured usage figure.

The brief

The person you brief is the person who solves it.

One conversation · 30 minutes · no monologue