The AAV-ML field is moving fast. Understandably, many of the claims being made in it do not hold up to close reading. And most programs, at some point, need a specific computational problem assessed or solved.
Advisory & Consultation covers both. Strategic guidance for organizations navigating ML investment decisions, partner relationships, or program direction. And bounded project engagements for programs that know what they need and want it done by someone who has built these models before.
Executive & Investor Advisory
Independent technical opinion for decision-makers evaluating AAV-ML platforms, collaborations, or acquisitions. Covers vendor and platform due diligence, collaboration structure review, experimental data asset assessment and investment evaluation.
For when you need someone with no stake in the outcome who can read what the ML claims actually mean and tell you whether they hold up before you commit resources to finding out the hard way.
Technical Advisory
Strategic guidance for programs building or running AAV-ML capability. Covers roadmapping, experimental design review, partner assessment, and program direction.
For organizations that need scientific direction from someone who has built this capability before, not someone learning the problem alongside them.
PROJECT-BASED CONSULTATION
Some programs know what they need. They have data, a target, a specific computational question, and a timeline. They need rigorous ML applied to a defined problem, delivered by someone who understands both the biology and the models well enough to know where each one breaks.
Bounded engagements around a specific, defined problem. Defined scope, timeline, and deliverable.
Examples
You have screening data (fitness scores, functional readouts, selection results, etc) and want to know which sequences to prioritize or which variants are worth synthesizing next.
You are planning a library and want the design to reflect what your prior screens already found.
You have a specific capsid property to optimize and need a scoped computational engagement.
You are evaluating capsid candidates from an external source and need an independent computational opinion before committing resources.
You have an ML question your internal team cannot answer. Not a general capability gap, but a defined problem with a defined answer. And need someone who has built these models before not to take chances.
PROGRAM AUDIT
For programs that have invested in computational capability but whose results are not matching the investment. Hit rates plateau. Models stop improving. The team cannot determine from the inside whether the ceiling is in the library design, the ML pipeline, the data itself, or the biology.
A Program Audit is a direct engagement with your program's actual work. I go into your data, your pipelines, your modeling history, and your experimental validation record and form a specific opinion about where the constraint is.
The deliverable is a written diagnostic: what is limiting your program computationally, what is fixable, and what is not.
AUDIT-LED REMEDIATION
When a diagnostic identifies a fixable problem, the next question is execution. What that looks like depends entirely on what the audit found. It may be a library designed computationally to reach the regions of sequence space the current approach has not sampled. It may be a modeling approach that extracts signal the current pipeline is discarding. It may be a targeted fix to one component, or a coordinated redesign of several together.
The scope is defined by the diagnostic, not by a predetermined service package. The engagement begins when the audit has established what is needed and why.