DingAI · Museums

Artifact Recognition Platform

2023 · Founding ML Engineer · Co-founder

An AI image classification platform identifying museum artifacts, built with two co-founders and external partners.

Composition

  • InferenceDocker · AWS Lambda
  • Service layerFlask · FastAPI
  • ConcurrencyPython multithreading
  • PersistenceManaged database
  • ValidationStress + reliability testing

Mission & Constraints

  1. Co-led product direction and system architecture — roadmap and deployment strategy alongside the build, not after it.

  2. The pipeline was containerised and serverless from the start, which is where the 90 % operational cost reduction came from; portability was the design goal and the cost was the consequence.

  3. Performance work was measured, not assumed: multithreaded processing and stress-tested tuning in Python produced a 94 % speed improvement, with the bottlenecks found by load testing rather than by inspection.

Implementation

Inference

Docker · AWS Lambda

Service layer

Flask · FastAPI

Concurrency

Python multithreading

Persistence

Managed database

Validation

Stress + reliability testing

Measured Outcomes

Operational infrastructure cost

−90 %

baseline toresidual

DingAI — serverless AI image classification

Inference pipeline speed

+94 %

baseline toachieved

DingAI — multithreaded processing, stress-tested in Python

Questions about this work?

Happy to go deeper on the architecture, the constraints, or how the numbers were measured.