We build secure, auditable AI systems for investment teams, connected to the documents, models and tools they already use.
Hold toblast
Systems that act rather than answer, evaluated continuously, handed over with the keys.
Fenro is an applied AI engineering practice founded by Alex Feng and Philip Rocki. The name is the two surnames, which is also roughly how the company works: there is no account layer between you and the people writing the code.
More about usFocused systems.
Measured execution.
From workflow to outcome
The practice
From the first auditto a system in production.
Across private markets, teams still move between documents, Excel, internal systems and recurring manual checks.
We connect those steps into secure AI workflows that can extract information, run controlled calculations, update models, surface risks and keep people in control of material decisions.
Reads across deal documents, surfaces key facts and risks, and keeps every answer tied to its source.
Extracts financial inputs, validates them and feeds structured data into the models your team already uses.
Tracks recurring reporting, recalculates key metrics and surfaces changes that require analyst attention.
From claim to proof
A solution is a claim.A system is the proof.
Selected work
One trusted answer across every regional office, grounded in approved documentation and linked back to its source.
View case studyInvestment teams
Document-heavy investment workflows, connected to the tools, data and controls already in place.
Underwriting, covenants and portfolio monitoring
Diligence, investment analysis and portfolio workflows
Research, reporting and investment operations
Data rooms, modelling and transaction execution
Due diligence, asset monitoring and reporting
Credit analysis, restructurings and complex documentation
Investment research, deal evaluation and portfolio knowledge
Reporting, document workflows and recurring controls
Where possible, we deploy into the client’s existing cloud environment, typically Microsoft Azure. Production resources, identity, data access and permissions can remain under the client’s control rather than being moved into a separate Fenro environment.
The architecture is designed to minimize unnecessary data movement. Depending on the client’s requirements, models, retrieval, storage and application components can be configured around their existing Azure environment, private networking and approved services. We define the exact data boundary before implementation.
For enterprise deployments using approved Microsoft/Azure model services, customer prompts and business data are not used to train the underlying foundation models by default. The exact setup is confirmed for each deployment.
Yes. We can build around Microsoft 365, Azure, Entra ID, SharePoint, Excel and Copilot rather than introducing an entirely separate stack. The goal is to extend the environment your team already trusts and uses.
Access can be tied to your existing Microsoft identity and permissions through Entra ID and role-based access controls. Users should only be able to retrieve documents, deals or portfolio information they are already authorised to access.
Where required, sensitive Azure services can be connected through private networking, private endpoints and controlled outbound access. The network design depends on the client’s security requirements and existing architecture.
We do not rely on the language model for everything. Retrieval is grounded in approved source material, important calculations are handled by deterministic code, and material actions can require human approval. We also test workflows against real cases before production.
Yes. Systems can be designed to log the user, source documents, retrieved evidence, model or workflow version, tool calls, calculations and approval steps. This creates an audit trail rather than a black-box chatbot.
Only to the extent explicitly agreed and required. We prefer least-privilege access, separated development and production environments, and client-controlled deployment processes. In stricter setups, production can be deployed through the client’s own CI/CD process without broad Fenro access.
No. The point is usually the opposite. We build around the systems already in place, such as Microsoft 365, SharePoint, Excel, internal databases, data rooms and existing investment workflows.
The workflow should be designed to surface uncertainty rather than hide it. Low-confidence outputs, missing evidence or material exceptions can be routed to a person for review.
The architecture can be designed around GDPR and Swiss data protection requirements, but compliance depends on the exact data, use case, region, vendors and client policies. We agree those requirements before production deployment rather than making a blanket compliance claim.
Fenro