I validate the technology in a target before you commit
and the technical integration plan before day one.
Manufacturing portfolio companies carry measurable operational inefficiencies: quality escape costs, unplanned downtime, yield variance. The challenge is knowing where AI can recover margin — and with what confidence — before allocating budget. A structured engagement: rapid data screening across all portfolio companies, on-site assessment on the top candidates, and a business case for each viable opportunity with estimated EBITDA impact, investment required, and implementation timeline.
This applies to any product with software inside that connects to a network: industrial controllers, edge gateways, smart meters, PLC systems, IoT sensors, building automation devices. If a known-exploited vulnerability is found in a dependency your product ships — a Python library, a firmware component, an embedded OS package — the 24-hour clock starts the moment you become aware. Missing it carries fines up to €15M or 2.5% of global turnover. Most manufacturers have no process in place. I've automated the entire workflow: daily dependency scanning against live vulnerability databases, CISA KEV cross-reference, and a pre-filled ENISA report ready when the clock starts. For PE/VC portfolios with connected-product companies, this is an unpriced compliance risk sitting in the portfolio today.
Independent technical validation on a target or add-on, structured for deal teams who need a defensible answer, not a longer report.
Fictional target, illustrative only — shows the structure and evidence standard of a real engagement.
A technical integration roadmap delivered before closing, so the platform and the add-on know what to expect from each other on day one.
Performance numbers, architecture decisions, or certification claims that the deal team can't independently assess — and that materially affect valuation or risk.
The target makes sense strategically, but no one has mapped what connecting it to the existing platform actually requires — until it becomes a day-one problem.
You need a defensible technical opinion fast, from someone who has built and deployed the systems in question — not just evaluated them on paper.
The diligence is informed by building, not just reviewing.
Independent statistical evidence packages for AI models in medical devices — dataset characterisation, stratified performance, calibration, drift detection, conformal prediction — structured for EU AI Act and MDR Annex II technical file submission. The same physics-calibrated measurement discipline applied to a different regulatory context.
See the service →An edge AI anomaly detection system — PatchCore on NVIDIA Jetson, industrial camera, edge-to-cloud telemetry — developed and operated as a production system, not a prototype. The research questions it raises about performance measurement, distributional shift, and uncertainty quantification are the same ones that appear in every DD and every regulatory validation we run.
See Inspector →A 30-minute call to see if there's a fit. No commitment, no proposal before we've understood the deal.