Real constraints. Real alpha. 10 pockets, one brain.

Turn heterogeneous inputs (decks, docs, calls, data-room, CRM exports) into a consistent schema—ready for scoring, constraints, and comparators.

Multi-signal screening across team, market, traction, economics, risk, and fit—calibrated thresholds + explainable decision gates.

Pre-check hard constraints before time is wasted: tickets, LPA caps, geo/sector exposure, ownership targets, concentration, conflicts, policy rules.

Selection under constraints + entry pricing discipline: walk-away thresholds, scenario consistency, sensitivity to assumptions, and portfolio impact.

Ticket sizing with concentration guardrails: per-line caps, top-k rules, risk budgets, tempered Kelly, CVaR sizing, and policy limits.

Detect hidden correlation clusters across sector, geography, macro drivers and factor exposures—so the portfolio isn’t “diversified on paper only”.

Tail risk control (CVaR / stress), hedging policies when relevant, and risk-to-KPI trade-offs across TVPI/DPI/IRR—not just “risk scores”.

Syndication mechanics: participation sizing, rights, seat allocation under oversubscription, co-invest allocations, and rule-based fairness options.

Time-phased deployment: windows, capital call smoothing, subscription line constraints, and cash-drag minimization through multi-period optimization.

Follow-on strategy under the power law: optimize the next euro per line (marginal MOIC), pro-rata rights usage, and opportunity costs.

DPI-aware decisions: partial secondaries thresholds, exit timing, liquidity plans by vintage, and KPI trade-offs under constraints.

Gross-to-net realism: fees, carry mechanics assumptions, tax impacts, and scenario deltas—so “optimal” doesn’t break when netted.

Signal harvesting + learning loops: recalibrate thresholds, learn which signals predict outcomes, and feed improvements into W2–W11.

• 0–1 & Multi-dimensional Knapsack (MDKP) — select deals under multi-constraints.
• Multiple-Choice Knapsack (MCKP) — exclusive choice among ticket/vehicle variants.
• Binary Quadratic Knapsack (BQKP) — interactions (synergies/antagonisms).
• Set packing/covering/partitioning — compatible subsets maximizing value.
• Conflict-graph knapsack — forbid incompatible combinations.
• Cardinality constraints — cap number of lines per pocket.
• Chance-constrained selection — satisfy constraints with target probability.

• Stable set / MIS — conflict-free portfolios.
• Cliques / coloring — concentration clusters & minimum diversification.
• Matching / assignment — allocate resources to opportunities.
• Dependency graphs — milestones prerequisites before follow-ons.

• Min-cost flow / transportation — move cash at lowest drag.
• Time-expanded networks — pacing across periods (call → invest → distribute).
• k-shortest paths / circulation — schedule trade-offs.
• Max-flow / min-cut — feasibility checks under capacities.

• RCPSP — resource-constrained scheduling for pacing/reserves.
• Lot-sizing — smooth calls vs fixed/variable costs.
• Cash-management — optimal buffers and drawdown policies.

• Markowitz (QP) — mean–variance under constraints.
• CVaR (LP) — tail-risk control via scenario linearization.
• Risk parity / budgeting — balance risk contributions.
• Robust optimization — stability under uncertainty.
• Stress testing — sector/geo/liquidity shocks.

• Multi-stage stochastic programming — decisions across scenarios over time.
• Dynamic allocation — progressive ticket & reserve choices.
• Optimal stopping — timing to sell/partially exit.
• MDP / light RL — pacing policies under constraints.

• Bandits — learn predictive signals, explore/exploit.
• Bayesian optimization — auto-tune thresholds & penalties.
• Causal inference / uplift — estimate follow-on counterfactual value.
• Calibration — well-calibrated scores for threshold decisions.

• Time series — calls & distributions forecasting.
• Survival analysis — time-to-round/exit by deal profile.
• GLM/GAM & Hierarchical Bayes — share information across pockets.
• EVT & copulas — heavy tails & cross-pocket dependencies.

• CP-SAT — validate thousands of rules (hard/soft).
• SAT/SMT checking — catch logical inconsistencies in policies.
• Explainability — attribute KPI deltas by lever/constraint.

• MILP / MIQP / MISOCP — industrial-grade branch-and-cut.
• Decomposition — Benders, Dantzig–Wolfe, column generation.
• Lagrangian relaxation — fast bounds & high-quality heuristics.
• Metaheuristics — VNS, GRASP, SA, GA, local search.
• Monte Carlo — outcome distributions & confidence intervals.