Education & Research
A doctorate in AI safety and alignment, a masters in computer science and a bachelors in robotics, alongside two working papers. The thread running through the research is evaluation validity, whether a benchmark measures the capability it claims to, and what follows when it does not.
Education
The route into this work runs from control systems and embedded engineering, through software and data platforms, to alignment research. Each step is visible in what I build now.
Paris American International University, France
AI safety and alignment.
Attenuation of Safety Alignment Across Language and Modality in Multimodal LLMs
University of East London, United Kingdom
AI, computer vision, advanced software engineering and big data analytics.
The software engineering half of the work, from system design through to the data platforms underneath.
Parul University, India
Robotics, control systems and embedded engineering.
Where the perception and on-device inference work comes from, including the drone vision and firefighting robot projects.
Doctoral research
Ph.D., Artificial Intelligence & Machine Learning
How much of the apparent cross-modal safety gap is a reading-ability artifact rather than an alignment gap?
Multimodal safety benchmarks report that guardrails weaken when a harmful request arrives as an image rather than as text, or in a lower-resource language rather than in English. That result is usually read as an alignment failure: the safety training did not transfer.
The measurement cannot distinguish two very different models. One understood the request and refused. The other could not parse it at all. Both score as a refusal, and both score as a safety pass. A model that simply cannot read Swahili looks as aligned as one that read the request and declined it. The apparent gap may be measuring comprehension, not alignment.
A comprehension control run alongside the safety evaluation. Before a refusal counts as evidence of alignment, the model has to demonstrate on a benign task that it could read the input at all. Separating the two lets the cross-modal gap be split into the part that is an alignment failure and the part that is a perception artifact.
Status
Active research. The instrument and the controls are designed; the measurement has not been run, so no results are reported.
Release policy
Prompts come only from already-published benchmarks, with no novel attacks. No working jailbreak strings are committed, and results are reported as aggregate rates rather than per-prompt transcripts.
Working papers
Both are in preparation rather than published, and are labelled as such here for the same reason every figure on this site cites its source.
Failure modes of tool-augmented agents measured per topology: loops, tool misselection, error cascades and context exhaustion, each with a reproduction seed and a rate. The finding that a supervisor topology can cost several times the tokens for no accuracy gain came out of this work.
mcp-server-and-agentThe dissertation's central argument: that cross-modal and cross-lingual safety benchmarks conflate refusal with comprehension, and that a comprehension control is required before a refusal rate can be read as an alignment result.
Instrument and controls designed; measurement not yet run.
Agenda
Whether a benchmark measures the capability it claims to. Judge calibration, comprehension controls, contamination, and the smallest delta a sample size can actually detect.
Preference-trained models learn to agree rather than to be right, because agreement is what the reward signal rewards. Measuring sycophancy separately from accuracy, and finding where an objective is being gamed rather than satisfied.
How alignment behaves away from the language and modality it was trained on, and how much of the observed change is a measurement artifact rather than a real gap.
Failure modes of tool-using systems as a measurable property of topology, covering loops, misselection, error cascades and context exhaustion, and where human review belongs in an automated pipeline.
Benchmarks lose meaning once they are in the training set. Decontamination that is checked rather than assumed, and held-out construction that survives the next scrape.
Confidence gating as a design problem: where the threshold goes, what it costs per class, and how to tell an abstention apart from a failure.
Contact
Happy to talk about anything here, or about LLM systems, evaluation and serving generally. Mentioning a repository by name gets you a faster and more useful answer.