AI Marathon Coach Grounded in 90 Days of Actual Training Data

The Problem

Generic marathon training plans are static artifacts that do not adapt to current readiness, recovery, or long-run recency.

Data-heavy platforms can show ATL, CTL, and TSB trends, but athletes still need to interpret those signals and translate them into concrete next-step training decisions.

For marathon preparation, guidance had to reflect the last 90 days of actual workload rather than assumptions about the athlete's baseline fitness.

The Approach

Integrated Garmin Connect via OAuth 2.0 and ingested activity history, heart rate, and daily stats directly from the API.

Built a dashboard around endurance-critical signals: weekly mileage on a 12-week rolling basis, ATL/CTL/TSB load balance, zone distribution across Zones 1-5, and long-run history over 10 miles.

Implemented a multi-turn AI coaching interface that passes the previous 90 days of Garmin data as grounded context to Claude via the Anthropic API (with Bedrock failover).

Ensured follow-up questions retain context so recommendations account for current TSB, recent intensity distribution, and long-run spacing.

The Impact

• Authenticated product ingests real Garmin data — activity, heart rate, and daily load — via a scheduled poller plus on-demand sync, then grounds every recommendation in the athlete's last 90 days; the prompt forbids inventing any mileage or pace not in the data
• A public no-login demo runs the coaching interface against a realistic sample athlete, so anyone can try the experience without connecting an account
• Surfaces trend-level signals — e.g. unintended intensity drift despite stable weekly mileage — from ATL/CTL/TSB and zone distribution rather than generic plan templates
• Same exposure-safety controls as the finance tool: a hard daily cost ceiling, per-session rate caps, and a default-off kill switch