Lead ML Architect
SKILLS
FULL DESCRIPTION
Lead ML Architect
Company: [Employer hidden — sign up to reveal]
Location: London (On-site)
Job Type: Full-time
Experience Level: Mid-Senior level
About the Role
We're looking for a Lead ML Architect to build the core intelligence engine for [Employer hidden — sign up to reveal]'s vulnerability correlation platform. This isn't about wrapping LLMs around existing tools - it's about architecting a hybrid system where deterministic logic (entity resolution, feature engineering, hard deduplication) provides the scaffolding and LLMs add semantic reasoning on top. You'll work directly with the founder to translate security heuristics into scalable, explainable pipelines that turn 100,000+ noisy alerts into 5-10 actionable Focus Areas. This is an in-person role based in London.
What We're Looking For
Hybrid ML Architecture
We need someone who understands the ML/AI landscape well enough to make smart decisions about what to build vs. what to use. This is about creativity, fast experimentation, measurement, and knowing when to leverage existing solutions rather than building everything in-house. Less hardcore math, more pragmatic product sense.
Excellent Fit
- Shipped ML/AI products by smart integration of existing tools and services
- Strong understanding of the ML ecosystem-knows what's possible, what tools exist, and how to evaluate them
- Track record of fast experimentation and iteration with measurement driving decisions
Good Fit
- Understands the ML/AI landscape enough to have informed opinions on what to use vs. build
- Experience shipping features quickly by combining existing tools and services
- Comfortable learning new tools/approaches and evaluating them pragmatically
- Has built data pipelines or systems that process/transform data at scale
Borderline
- Primarily LLM prompt engineering without deep systems/feature engineering experience
- ML experience focused on standard supervised learning without hybrid architecture complexity
Not a Fit
- Only LLM wrapper projects with no deterministic engineering layer
- Purely theoretical ML background without production systems experience
Security Domain Fluency
Surface-level understanding of security concepts is helpful but not essential- we can teach the domain knowledge. What matters more is the ability to think abstractly about complex data problems. That said, candidates with prior exposure to vulnerability data, infrastructure concepts, or security operations will ramp faster.
Excellent Fit
- Direct experience with vulnerability management, bug bounty, or security operations
- Can articulate how attackers chain vulnerabilities to reach critical assets
Good Fit
- Familiarity with security concepts (CVEs, CVSS, attack surfaces)
- Experience with compliance or infrastructure security tooling
Borderline
- No security background but strong interest and fast learner
Not a Fit
- No interest in learning security domain
Pragmatic Builder DNA
We need someone with the 'Phase 0 first' mindset- starting with deterministic deduplication before touching LLMs, building evaluation harnesses before declaring victory, understanding why RAG isn't always the answer for structured knowledge. This is about engineering discipline and pragmatism over hype.
Excellent Fit
- Builds evaluation/measurement systems before shipping ML features
- History of choosing boring technology when appropriate
- Can articulate when NOT to use ML/AI for a given problem
Good Fit
- Values iteration speed and measurement over perfect architecture
- Experience debugging and improving production ML systems
- Comfortable with ambiguity and changing requirements
Borderline
- Tendency to over-engineer or reach for complex solutions first
- Limited experience with production system constraints
Not a Fit
- Builds features without success metrics or evaluation
- Follows hype cycles without critical evaluation
Similar Problem Domains
Experience at companies dealing with similar challenges is valuable-noisy data that needs aggregation, context building from disparate sources, signal extraction from high-volume streams. This translates directly to understanding our core problems even without security background.
Excellent Fit
- Built systems processing 100K+ events into prioritized insights
- Experience with entity resolution or record linkage at scale
- Worked on alert correlation, deduplication, or noise reduction
Good Fit
- Experience with high-volume data processing pipelines
- Built systems that aggregate or correlate data from multiple sources
- Familiarity with graph-based or relationship-heavy data models
Borderline
- Data engineering background without ML/correlation focus
Not a Fit
- Only worked with clean, structured datasets
Application Instructions
Apply for this role via email to [Employer hidden — sign up to reveal] with subject 'Lead ML Architect Application'.