Hiring 4 Full-Stack AI Engineers: How a Growth-stage PropTech in Paris Scaled with Second Talent

The Challenge

The company had reached $7M in annual recurring revenue by early 2025. Their platform helped real estate professionals assess property values, but the process relied heavily on manual analysis. Users spent an average of 23 minutes per property valuation. Competitors were launching AI-powered tools that cut that time to under 5 minutes.

The technical challenge was finding developers who could handle both sides of the equation. They needed engineers who understood machine learning model development and could also build production-ready React and Node.js applications. Splitting the work between an ML team and a frontend team would add 4 to 6 weeks of coordination overhead. The company had tried this approach on a previous project and missed their launch window by 11 weeks.

Hiring in Paris presented two problems. First, the talent pool for full-stack AI engineers was limited to about 40 qualified candidates, most already employed at large tech companies. Second, the salary expectations ranged from €95,000 to €130,000 per developer annually. At their current burn rate, adding four developers at that cost would push their runway from 18 months down to 11 months.

They needed a team that could start immediately, work across the full stack, and fit within a budget of $25,000 per month for all four developers. Local recruitment agencies quoted 6 to 8 week timelines just to present the first candidates. The company had 12 weeks to ship a working product before their largest client renewed or switched to a competitor with AI capabilities.

The Solution

Second Talent presented 12 candidates within 9 days. All candidates had 6+ years of professional experience. Each had shipped at least two production AI features in the past 18 months. The technical evaluation focused on three areas: building and deploying ML models, React performance optimization, and API design for real-time data processing.

The company selected 4 engineers, all based in the Philippines. Three had backgrounds in fintech where they built fraud detection and risk assessment models. One had worked on computer vision applications for construction tech. All four were AI-native developers who used tools like GitHub Copilot, Cursor, and Claude for at least 40% of their daily coding work. This meant they could move faster than traditional developers on both model experimentation and frontend implementation.

Second Talent handled the EOR setup in the Philippines. The team was legally employed within 6 days. Payroll, benefits, tax compliance, and local labor law requirements were managed end-to-end. The company paid a single monthly invoice covering all four developers plus the EOR service fee. No need to establish a local entity or navigate Philippine employment regulations.

The developers started on March 3, 2025. They worked in a timezone 7 hours ahead of Paris, which created a natural handoff workflow. The Paris-based product team would define requirements and review work in the afternoon. The Philippine team would build and test overnight. Morning standups happened at 9 AM Paris time, 4 PM Manila time. This overlap window gave them 3 hours of real-time collaboration daily. The rest ran asynchronously, which actually accelerated the development cycle because there was always progress happening.

The Results

The AI valuation tool went live on April 28, 2025. Total development time was 8 weeks. The tool used a gradient boosting model trained on 340,000 historical property transactions. The React frontend displayed confidence intervals, comparable properties, and market trend data. Average valuation time dropped from 23 minutes to 4 minutes. User engagement, measured by daily active users completing valuations, increased 45% in the first month after launch.

The cost structure delivered significant savings. Each developer earned $6,200 per month. Total monthly cost for four developers plus EOR fees came to $26,300. Hiring the same team in Paris would have cost approximately €75,000 per month at the lower end of market rates. That translated to 65% cost savings while actually reducing time to market by an estimated 6 weeks compared to the coordination overhead of separate ML and frontend teams.

The AI-native approach showed up in the velocity metrics. The team averaged 127 pull requests per week. Code review cycles took 2.1 hours on average because AI tools caught most syntax and logic errors before human review. The developers used AI pair programming for 34% of the codebase, which meant faster iteration on model architecture and UI components. Technical debt remained low because AI-assisted refactoring happened continuously rather than in dedicated sprints.

“We needed people who could train a model in the morning and build the UI for it in the afternoon. Second Talent found us four engineers who did exactly that. Eight weeks from kickoff to production. Our engagement metrics jumped 45% in the first month. The EOR setup meant we focused on building product instead of navigating foreign employment law.”

CTO