Most recruiting tools find who's available. NCompas finds who's right — technically skilled, culturally aligned, and likely to stay.
Our AI is built in-house — not licensed from a vendor — trained on real placement outcomes across technical depth, cultural fit, and retention probability. Evidence-backed shortlists, fully explainable, no keyword matching.
Why NCompas
The difference between a recruiter who can read a GitHub profile and one who can't is the difference between finding the right ML engineer in 6 weeks and spending 4 months forwarding CVs of data analysts who've done a Coursera course.
Our recruiters understand the difference between a data engineer and a data scientist, between a .NET architect and a .NET developer, between a machine learning engineer and a data scientist who's read a few papers. We read GitHub profiles, assess code quality, and understand why a candidate's arXiv paper on causal inference matters for your ML team. This is not a CV-forwarding operation.
The AI recruitment tools we use are tools NCompas has built — not licensed from a third-party SaaS vendor. Our sourcing engine, LLM screening model, technical assessment platform, and predictive retention model are proprietary. We can explain exactly how every score is computed, what evidence it's based on, and why a candidate ranked where they ranked. No black-box vendor magic.
The best candidates are not on job boards. They're employed, doing excellent work, and not actively looking — which is precisely why they're excellent. Our multi-source AI discovery engine finds them via their public technical footprint: GitHub contributions, arXiv papers, conference talks, Stack Overflow answers. 65% of our successful placements are candidates who never applied anywhere.
Salary benchmarks are not from an annual survey conducted 8 months ago. Our talent intelligence platform has real-time data on what similar roles are offering today, in your geography, for your skill requirements. We tell clients when their budget is misaligned with the market before they waste months of process on candidates who will ultimately decline. Unpleasant news delivered early is better than rejection delivered late.
A hire who leaves at 9 months cost you 1.5–3× their annual salary in recruitment, onboarding, lost productivity, and re-hiring costs. Our predictive retention model is why our 12-month retention rate is 91% — 20 percentage points above the industry average. We optimise for the 12-month outcome, not the placement fee.
Every candidate who reaches shortlist receives a structured report: 12 capability dimensions scored with evidence, technical assessment results with annotated evaluation, predictive success probability with confidence interval, and market context (how this candidate compares to the 14 other pipeline candidates we've assessed this month for similar roles). Hiring decisions backed by data, not vibes.
Six AI recruitment capabilities
Six interconnected AI capabilities that transform every stage of the talent acquisition pipeline — so the right engineers are found, fairly assessed, and stay.
The best engineers for your roles aren't on job boards — they're on GitHub maintaining open-source libraries, on arXiv publishing ML research, speaking at KubeCon, and on LinkedIn with "Open to work" set to private because they get 40 irrelevant InMail messages per week. Our AI sourcing engine crawls 14 professional signals simultaneously — code commits, publications, conference presentations, Stack Overflow reputation, patent filings, open-source contributions — to build a richer capability picture than any CV provides.
The best engineers for your roles aren't on job boards — they're on GitHub maintaining open-source libraries, on arXiv publishing ML research, speaking at KubeCon, and on LinkedIn with "Open to work" set to private because they get 40 irrelevant InMail messages per week. Our AI sourcing engine crawls 14 professional signals simultaneously — code commits, publications, conference presentations, Stack Overflow reputation, patent filings, open-source contributions — to build a richer capability picture than any CV provides.
Traditional ATS keyword matching rejects the Kubernetes maintainer because their CV says "container orchestration" not "Kubernetes". Our LLM-powered resume intelligence understands semantic equivalence, infers capability from project descriptions, weights recency of skills, identifies skill trajectories, and matches against a structured role definition — not a keyword list. Every candidate receives a structured 12-dimension capability assessment with evidence citations.
Take-home coding tests lose 40% of senior candidates who have jobs and no time. Live whiteboard sessions measure nerves, not capability. Our AI-powered technical assessment platform uses real-world scenario challenges evaluated by LLM + rubric, structured technical conversation with AI analysis, and code review exercises assessed for architecture quality — not just correctness. Senior engineers spend 60–90 minutes on realistic work, not arbitrary puzzles.
Hiring the technically-correct candidate who leaves in 6 months costs 1.5–3× their annual salary (SHRM). Our ML model, trained on 2,400+ placed candidates and their 12-month outcomes, predicts success probability across four dimensions: technical growth trajectory, team collaboration pattern match, role complexity alignment, and cultural value fit — so hiring decisions account for likely retention, not just day-one capability.
Unstructured interviews are scientifically poor predictors of performance (r=0.28) and systematically favour candidates who went to the right schools or spoke with the interviewer's accent. We design structured competency-based interview frameworks, interviewer calibration programmes, blind CV review processes, and diverse panel composition — combined with AI-generated structured scorecards that require evidence for every rating. Measurably fairer, measurably better.
Real-time talent market data: which AI/ML skills are in surplus (oversupply → faster hiring, lower salaries), which are scarce (Python MLOps engineers in Manchester — 4 available, 140 open roles), where your competitors are hiring from, and what compensation benchmarks are moving. Workforce planning models for 12-month headcount growth scenarios — built before the board approves the headcount, not after.
Technology disciplines
Demand index reflects current open roles vs. available active candidates in the UK market — a higher index means longer hiring cycles and stronger compensation expectations.
Delivery approach
Five stages from role briefing to placed and thriving — with market intelligence at the start that prevents salary-mismatch rejections that stall 1 in 3 final-stage processes. Junior roles placed in as few as 2–3 weeks; senior and niche roles vary by complexity — we give you a realistic estimate upfront, not a promise we might miss.
We start with a structured role definition session covering technical requirements, team dynamics, growth trajectory, and success definition at 90 days and 12 months. You receive a talent market intelligence brief — real-time supply/demand analysis, salary benchmark for the specific skill set and geography, and an honest assessment of time-to-fill at the proposed compensation. No surprises at the offer stage.
Our AI sourcing engine deploys across 14 professional signals — GitHub, GitLab, arXiv, Google Scholar, Stack Overflow, LinkedIn, conference speaker directories, open-source contribution lists, podcast appearances, and published case studies. Passive candidate outreach sequences are personalised by AI using each candidate's specific technical footprint, building a qualified long-list fast.
LLM-powered resume intelligence across the long-list — semantic skill extraction, trajectory analysis, project-derived capability inference. 12-dimension structured scoring with evidence citations. AI technical pre-assessment (60-minute real-world scenario) deployed to top candidates. A shortlist of 6–10 with full structured reports follows — timelines shared transparently at kickoff.
Structured competency interview framework, evidence-based scorecards, interviewer calibration session, and diverse panel composition guidance. AI-generated interview guides tailored to each candidate's specific profile gaps. AI-assisted debrief facilitation ensures panel decisions are evidence-based, not post-hoc rationalisation of gut feel.
Offer construction using live market intelligence, equity-adjusted total compensation modelling (where applicable), and counter-offer probability assessment. Structured negotiation support. Post-acceptance: 90-day onboarding success programme with monthly check-ins, early flight-risk signal monitoring, and manager coaching for the critical first 30 days. Retention is not accidental.
Client outcomes
Four organisations where AI recruitment turned hiring blockages into built teams — with retention rates that make every placement a compounding asset.
The Challenge
Series B fintech with £28M raised needing to grow from 12 to 55 engineers in 12 months — with a specific requirement for 8 ML engineers who understood financial time-series modelling, 14 senior .NET engineers for their trading platform, and a VP Engineering who could lead the London engineering function. Existing internal recruiter had been trying for 4 months with 0 ML engineer offers made.
What We Delivered
AI multi-source sourcing across GitHub (financial ML repositories), arXiv (quantitative finance papers), and LinkedIn. LLM resume intelligence to identify .NET engineers with financial domain experience from their commit histories and project descriptions. Structured technical assessment calibrated to the firm's actual codebase complexity. Talent market intelligence confirming that the offered salary for ML engineers was 18% below market — renegotiated upward with data before any process started.
Series B fintech with £28M raised needing to grow from 12 to 55 engineers in 12 months — with a specific requirement for 8 ML engineers who understood financial time-series modelling, 14 senior .NET engineers for their trading platform, and a VP Engineering who could lead the London engineering function. Existing internal recruiter had been trying for 4 months with 0 ML engineer offers made.
NHS-partnered health tech company building an AI diagnostic platform — required FHIR-specialist engineers, clinical NLP data scientists, and a Head of AI with both ML depth and NHS regulatory experience. Regulatory constraints meant candidates needed UK-right-to-work and NHS DBS clearance, cutting the addressable talent pool to approximately 200 people nationally.
FTSE 250 retailer transforming their 80-person IT function from legacy .NET Framework and Oracle skills to cloud-native .NET 9, Azure, and AI — needing to hire 35 net-new cloud-native engineers while managing the sensitivity of existing staff whose roles were changing. Budget for the external hiring programme was fixed at £280K for the year.
AI product startup with 4 founders and a strong product but no engineering team — needed to build from 0 to 30 in 9 months for their Series A milestone, with specific requirements for frontier AI experience (GPT-4o fine-tuning, RAG systems, agentic frameworks) in engineers who would accept below-market cash in exchange for meaningful equity. Market sentiment for AI engineers in 2024 made this a highly competitive search.
average reduction in time-to-hire with AI-powered sourcing, screening, and scheduling — translating directly into months of lost engineering productivity recovered per role, compounded across every open requisition in a growing technology team.
of NCompas successful placements are passive candidates who never applied anywhere — found via their public technical footprint on GitHub, arXiv, and conference records, not via job boards where every recruiter is competing for the same 35% who are actively looking.
predictive validity of structured AI-assisted interview process — vs. r=0.28 for unstructured interviews. A hire on the basis of a gut-feel conversation is essentially a coin toss; structured evidence-based assessment is measurably predictive of on-the-job performance.
12-month retention rate across all NCompas technology placements — 20 percentage points above the LinkedIn industry benchmark of 71% for tech roles. A retained engineer compounds value; a churned one compounds cost.
Insights from real-world engineering, cloud, and AI leaders - helping you make better decisions, faster.
We're putting the finishing touches on this. Check back soon for in-depth insights.
Start with a free talent market intelligence brief — we'll run a real-time supply/demand analysis for your specific role, geography, and skill requirements, and give you an honest assessment of time-to-hire and salary benchmark before any search begins.