For PE, VC, and family offices in Geneva and Romandie.
IC memo automation, deal screening, diligence synthesis, and portfolio monitoring. Production-grade, auditable, source-cited. For finance teams and SME CFOs, productized AI on a fixed scope and fixed fee.
Investment research and reporting are still largely manual, fragmented, and hard to audit. We replace that work with structured systems that produce the same output every time.
By combining finance, software engineering, and Claude-based AI, we build tools that produce consistent, auditable outputs instead of one-off work. Every system ships with full test coverage and a Codex-audited review.
The hero offer is deal intelligence for PE, VC, and family offices. The mandate catalog publishes per-deliverable and retainer pricing for every recurring engagement. The supporting offer is productized financial analytics for finance teams and SME CFOs. All three are scoped tightly, implemented to test coverage, and deployed as production tools.
AI investment research and deal intelligence for funds and family offices in Geneva and Romandie. Scoped per mandate, test covered, deployed as production tools.
The full catalog. 13 priced mandates across per-deliverable packs, monthly retainers, quarterly reviews, and white-label terminal seats. Geneva-anchored CHF.
Productized AI for finance teams and SME CFOs. Fixed scope, fixed fee, fixed timeline, from document intake to management reporting and internal knowledge assistants.
These are not client engagements. They are reference systems built to demonstrate capability, and they are available as starting points for client engagements. Each maps to a shipped, test-covered system you can open and inspect.
A deterministic recommendation aggregator that turns filings, financials, valuation, and risk into an IC ready memo. The rating is computed and locked before the model writes any prose, so the narrative can never override the decision.
Percentile scoring blended with a simplified LBO IRR, debt capacity tiering, and red flag detection across a screening universe. Built to surface a short list with the reasoning attached, not a black box rank.
The full engine stack integrated into a Bloomberg style research terminal with institutional workspaces for live market, research, comps, and portfolio work. The capstone that the prior systems compose into.
These systems form the technical foundation of Frostaing AI's implementation capability. Each ships with full test coverage, a Codex-audited review, and a live app. The capstone integrates the prior ten into a single research terminal.
Ten systems collapse into one research desk.
Unified Bloomberg-style research terminal integrating all 10 prior engines into 68 institutional workspaces (Market, Research, Options, LBO, Comps, Portfolio, sector pages, intelligence surfaces). 7,734 backend tests passing, plus 2,542 frontend.
Six engines decide. The LLM only writes the memo.
Deterministic recommendation aggregator. Engines run in parallel. The model writes the prose. The rating is locked.
HRP with RMT cleaning beats classical Mean Variance.
Four optimizers, three covariance methods. Ten year rolling backtest across fourteen ETFs. Sharpe 0.62 at 10 percent target volatility.
Frostaing AI is a Geneva-based AI and finance systems firm building AI investment research and deal intelligence for private equity, venture capital, and family offices, with a supporting productized offer for finance teams and SME CFOs.
Founded and operated by Francois Rostaing, Claude Certified Architect (Anthropic), supported by a network of specialist collaborators.
Member of the Claude Partner Network (Anthropic).
The work is anchored in three signals: real financial understanding, strong data engineering, and decision tools investors and operators actually use. The eleven shipped systems are the firm's technical foundation; client engagements extend that stack into production research, automation, and decision-support infrastructure.
We work with PE, VC, and family office teams on deal intelligence, and with finance teams on productized analytics. A short scoping call is the fastest way to see if there is a fit.
Book a call