Case 04 — AI / SaaS / Scale-up

How we scaled an AI logistics team to serve 1,200+ enterprise clients — including Walmart, Falabella, and Unilever.

SimpliRoute was growing fast. Their engineering team couldn't keep up with the product roadmap. They needed senior engineers who could own complex modules from day one — not junior hires who needed 3 months to ramp.

Timeline overview

Month 1

2 senior engineers embedded in product squads

Onboarding in days, not weeks. Both engineers were contributing to production by day 5 — backend routes optimization and fleet monitoring dashboards.

Month 2–4

SAP, Oracle, WooCommerce integrations shipped

3 major enterprise integrations delivered — critical for landing Walmart and Falabella as clients. Each required custom field mapping, error handling, and retry logic at scale.

Month 5–8

Cold chain module + ADA AI agent features

Temperature monitoring for refrigerated logistics and automated incident detection within the ADA AI layer. High-complexity features shipped on schedule.

Ongoing

Long-term embedded partnership

FlyDevs engineers remained a core part of the SimpliRoute product team through the company's continued growth phase.

Key technical decisions

🔄

Built a unified integration adapter instead of one-off connectors

The first two integrations (SAP + Oracle) were built with a shared adapter pattern, making the third (WooCommerce) 40% faster to deliver. Slightly more upfront investment, significant long-term payoff.

🌡️

Chose MQTT over HTTP polling for cold chain monitoring

Temperature sensors send updates every 30 seconds. HTTP polling at that frequency at scale was untenable. MQTT pub/sub reduced server load by 70% compared to the prototype polling approach.

🤖

Rule-based anomaly detection before ML

The AI agent team wanted ML-based anomaly detection immediately. We pushed for a rule-based v1 first — it shipped in 3 weeks instead of 3 months and gave the team real production data to train the ML model properly.

What almost went wrong

⚠️

The Walmart integration had an undocumented field mapping edge case

Walmart's WMS used a non-standard address format for multi-stop routes. We caught a silent truncation bug in staging that would have corrupted delivery records for routes with more than 12 stops. Fixed before it ever touched production.

Result

Enterprise clients

1,200+

Visits optimized

347M+

KM planned

603M

Delivery rate

95.6%

Want us to do this for your product?

Book a free 20-minute call. We'll scope your project, recommend the right team composition, and give you a realistic timeline — no pitch, no pressure.

Book a Free Call →