Project
SmartShots
A web and Chrome-extension product for collecting, organizing, and acting on digital screenshots and web artifacts.
Year
2026
Status
active
Type
AI product
Role
Product engineer and builder
Highlights
- Designed capture and review workflows that connect browser context, screenshot metadata, and downstream AI analysis.
- Hardened storage and access paths across private assets, including migration work and route-level enforcement.
- Built admin and observability surfaces for monitoring users, event volume, error patterns, logs, subscriptions, email, storage, deletion requests, and utility workflows.
- Improved dashboard and editor ergonomics so saved artifacts can be reviewed, organized, and acted on.
Project images
Screenshots, diagrams, and supporting visuals.
Custom admin and observability dashboard for tracking product health, event volume, logging, errors, subscriptions, storage, and operational workflows.
The main product surface: searchable screenshot collections with visible tagging and source context.
Desktop sync flow for batching local screenshots into the SmartShots system.
Public product positioning and early architecture-style visual for multi-platform capture.
Lessons learned
What the project made clearer.
Admin tools are product infrastructure
The admin dashboard is not ornamental back office work. Logs, event rollups, error views, subscription state, email state, storage tools, and deletion workflows make it possible to operate the product with confidence while the user-facing surface is still moving quickly.
Preserve source context early
The most trustworthy labels and retrieval signals come from capture time. Browser URL metadata, page context, and user intent are stronger than trying to infer everything later from an image alone.
Private media needs product-level clarity
Storage decisions are not just infrastructure work. R2 paths, authenticated proxying, and ownership checks directly affect whether the product feels reliable and safe.
AI features work best when inspectable
The useful AI layer is not magic categorization. It is an assistive layer that explains, labels, and organizes artifacts while leaving enough context for the user to review and correct it.
Context
SmartShots started from a familiar pattern: useful digital fragments are easy to capture and hard to revisit. The product sits between a screenshot tool, a lightweight research archive, and an AI-assisted workspace for turning saved artifacts into memory and action.
Product Work
The work has included the web application, browser-extension capture flows, screenshot metadata handling, dashboard ergonomics, reminder surfaces, source labeling, and AI-assisted analysis. A recurring theme has been preserving trustworthy context from capture time rather than letting later interpretation overwrite it.
Admin and Observability
SmartShots also needed a real operating layer: a custom admin dashboard for understanding whether the product is healthy and where attention is needed. That work included event logging, OpenTelemetry-oriented instrumentation, error visibility, top-event rollups, hourly activity, user and subscription views, email state, storage tools, deletion-request handling, and utility workflows for support/debugging.
This part of the product matters because SmartShots spans several moving pieces: a web app, Chrome extension capture, upload and compression flows, private media storage, AI analysis, and user-facing organization. The admin surface gives those systems a shared operational view instead of leaving failures hidden in separate logs or provider dashboards.
Engineering Notes
The system has had to balance product speed with private media handling. Storage work included R2-backed asset paths, authenticated proxying, and careful enforcement around user-owned objects. On the product side, capture metadata and browser URLs became important signals for keeping the interface honest. On the operations side, the system needed traceable events and inspectable logs so upload failures, analysis issues, account actions, and storage behavior could be diagnosed without guessing.