The dirty secret of the outsourcing industry is that most vendors do not vet their engineers with anywhere near the rigor they claim. A quick resume scan, a 30-minute phone screen, and a cursory coding exercise is the standard process at many agencies. The result is predictable: clients receive developers who look good on paper but cannot perform in production. At S-Technology, we built our vetting pipeline to solve this specific problem. Only 15% of candidates who enter our pipeline make it into our active talent pool. Here is exactly how that pipeline works, stage by stage.
Stage 1: CV Parsing and Automated Screening
Every application enters our system through a structured intake process. We do not review raw resumes. Instead, our CV parser extracts and normalizes data into a standardized profile: technology stack (with years of production experience per technology), project complexity (measured by team size, user scale, and architectural patterns used), English proficiency indicators, and employment continuity. We reject approximately 60% of applicants at this stage based on minimum thresholds. Candidates need at least 2 years of professional experience, demonstrated work with at least one modern framework in production, and evidence of working on teams rather than exclusively solo projects. This automated screening saves our senior engineers from reviewing hundreds of unqualified profiles.
Stage 2: Gemini AI-Powered Technical Scoring
Candidates who pass initial screening enter our AI-assisted evaluation layer built on Google's Gemini API. The system analyzes each candidate's profile against a weighted scoring matrix that evaluates 12 dimensions including framework depth, system design exposure, testing practices, and open-source contributions. The AI does not make accept/reject decisions. Instead, it generates a detailed technical brief that highlights the candidate's strongest areas, flags potential gaps, and assigns a preliminary score between 0 and 100. Our engineering reviewers use this brief as a starting point for the next stage. The AI layer reduces evaluation time from 45 minutes to 10 minutes per candidate while improving consistency across reviewers.
Stage 3: GitHub and Portfolio Verification
We verify every claim on a candidate's profile against their actual work. For developers with public GitHub profiles, we analyze commit history patterns, code quality in recent repositories, PR review habits, and documentation practices. We are not looking for green squares. We are looking for evidence of professional-grade engineering: meaningful commit messages, clean branching strategies, proper error handling, and test coverage. For candidates whose work is primarily in private repositories, we review portfolio projects and request code samples from past work (with appropriate permissions). Approximately 20% of candidates who passed Stage 2 are rejected here because their claimed experience does not match their demonstrated output.
Stage 4: Live Technical Interview
This is where the real differentiation happens. Candidates sit for a 90-minute live session with one of our senior engineers. The first 30 minutes cover system design: candidates are asked to architect a real-world system, such as a notification service handling 10,000 messages per second or a payment reconciliation pipeline. We evaluate not just the design itself but how the candidate communicates trade-offs, asks clarifying questions, and responds to constraints introduced mid-discussion. The next 40 minutes are a pair programming exercise on a production-grade problem, not an algorithm puzzle. We want to see how candidates write code in conditions that mirror actual project work: reading existing code, handling ambiguous requirements, and making pragmatic decisions under time pressure. The final 20 minutes assess communication and collaboration skills through scenario-based questions about handling code review disagreements, estimating work, and managing blockers on distributed teams.
The Power Ranking System: S Through D Tiers
Every candidate who completes all four stages receives a composite power ranking on a 5-tier scale: S, A, B, C, and D. S-tier engineers are exceptional across all dimensions: deep technical skills, strong communication, proven track record on complex projects, and high autonomy. They represent roughly 3% of candidates who make it through the full pipeline. A-tier engineers are strong generalists with at least one area of deep specialization. B-tier engineers are solid mid-level developers who perform well with clear direction. We only place S, A, and B-tier candidates with clients. C and D tiers are not placed. The ranking is transparent: clients see the breakdown of each candidate's scores across all evaluation dimensions, so they can make informed decisions based on their specific requirements rather than trusting a vendor's subjective recommendation.
Continuous Performance Monitoring
The pipeline does not end at placement. Every engineer in our talent pool is subject to ongoing performance tracking based on client feedback, sprint velocity data, and quarterly technical assessments. Engineers who consistently receive strong client reviews move up in ranking. Engineers who underperform receive targeted coaching, and if performance does not improve within 60 days, they are removed from the active pool. This creates a feedback loop that continuously improves the quality of our talent network. The result is measurable: our client retention rate across engagements exceeding 6 months is 94%, and the average time to replace an engineer who does not work out is 10 business days. Rigorous vetting upfront is what makes those numbers possible.