All projects

03 / ENGINEERING CASE STUDY · Products

LiftCircle

A native SwiftUI training app that combines workout logging with a Python retrieval-augmented coaching backend.

SwiftUIPythonRAGFirebaseNext.js
July 2025 — January 2026
Visit website

01 / CONTEXT

The problem

Useful training advice depends on workout context. LiftCircle brings logging, including reps in reserve (RIR), together with personalized coaching, while also handling the everyday product requirements of authentication and subscriptions.

02 / SYSTEM DESIGN

Architecture

From workout context to personalized coaching

SwiftUI app · workout logs + RIR
Python coaching backend · RAG
Semantic chunking
Retrieved context
Workout analysis + coaching
  • Firebase · authentication
  • StoreKit · subscriptions
  • Gold dataset · coaching evaluation
Conceptual view of the components described in the résumé, rather than a verified deployment topology.

03 / TRADEOFFS

Engineering decisions

01

Use retrieval to inform coaching

The backend uses retrieval-augmented generation and semantic chunking for workout analysis. Retrieval gives the coaching system relevant context to work from; output quality still requires evaluation rather than assuming that retrieved content guarantees a good recommendation.

02

Evaluate the backend as part of the product

The project includes gold dataset evaluation for the coaching workflow. Dataset composition, scoring criteria, and evaluation results are not published here, so this case study describes the evaluation approach without claiming a quality score.

03

Keep the training workflow native

SwiftUI brings workout logging and RIR input into the iOS app, alongside authentication and StoreKit subscriptions. These integrations make the coaching workflow part of a usable application rather than a standalone chat demo.

04 / VALIDATION

Evidence & measurement

This case study is based on the supplied résumé and the linked product website. Implementation details beyond that description, usage figures, and measured coaching outcomes are not publicly established here.

RAG coaching

The résumé describes a Python backend with semantic chunking and gold dataset evaluation for workout analysis.

View supporting reference

Native iOS experience

The résumé lists SwiftUI, RIR logging, authentication, and StoreKit subscriptions. The product website provides the public product reference.

View supporting reference

05 / NEXT ITERATION

What I’d improve next

Document coaching evaluation

Publish a representative, anonymized evaluation set, scoring criteria, and failure examples so readers can assess recommendation quality.

Measure retrieval separately

Compare retrieval approaches before evaluating generated answers. This would help distinguish missing context from reasoning or generation errors.

Project scope comes from the September 2026 résumé. The architecture is conceptual; the next steps are proposed improvements, not completed work.

NEXT CASE STUDYAutonomous Billing Proxy