Machine learning
PhD in Machine Learning (UK). A research and engineering career spent building and evaluating ML systems, and understanding precisely where their outputs can and cannot be trusted.
About LuSarv
Making AI accountable in the settings where it matters most
LuSarv is a London-based AI research start-up and consulting practice with a single, bold mission: accountable AI for high-stakes environments, designed for scrutiny from the outset.
The name
Lucent and Sarv
Lu, from Lucent, represents the source of light we want to bring to the AI industry, making AI outcomes and decisions explainable, accountable and traceable.
Sarv is the Persian name for the cypress tree, a symbol of eternal life, resilience, freedom and grace. It is the standard we build to: systems intended to endure scrutiny, not evade it.
The founders
Two disciplines, one conviction
LuSarv was founded in 2026 by two PhDs who bring together over thirteen years of post-PhD experience across academia and industry.
Human factors
PhD in Human Factors (US). Deep expertise in how people actually make decisions with technology in safety-critical settings, and what those systems owe the humans who rely on them.
Between them, the founders have collaborated with organisations including University of Leeds, Nissan, Financial Times, Amazon AWS and UK Department for Work and Pensions.
Core values
What we hold ourselves to
These commitments are not marketing language; they are the criteria against which our own work is judged.
01
Accountability over confidence
A high-confidence score is not accountability; a verifiable, immutable evidence trail is. We build systems that can stand in court, not just in benchmarks.
02
Evidence before assertion
No claim reaches a user without being grounded in verifiable source material. The LLM proposes; evidence decides. We treat generation as hypothesis, not as truth.
03
Safety by architecture, not by policy
Safety constraints must be structurally enforced, not written in guidelines and hoped for. Our designs make unsafe outputs architecturally impossible, not merely discouraged.
04
Sovereignty first
Sensitive data, such as patient records, legal files and government documents, must never cross an external boundary. We build AI that works entirely within your infrastructure, not despite it.
05
Rigour over heuristics
Confidence thresholds are not guarantees. We replace soft scores with mathematically certified uncertainty bounds, because in healthcare, “probably correct” is not good enough.
06
Transparency as default
Every output is traceable to its source. Every decision carries a timestamped, hashed audit log. Transparency is not a feature; it is the foundation we build on.
07
Human oversight, machine rigour
We do not aim to replace clinical or professional judgement. We build AI that gives practitioners verifiable, complete and accountable information, so the human who decides can trust what they are reading.