AI-Native Quant Hedge Fund

Each fund holds all constituent stocks at all times — there is no cash allocation. Stocks are classified into two tiers based on their AI-generated decision signal:

• Active positions ("Buy" signal): Weighted proportionally to their conviction scores, with a per-fund maximum weight cap. Excess weight above the cap is redistributed proportionally among uncapped active positions.
• Passive positions ("Hold" or "Sell" signal): The residual portfolio allocation (1 minus the sum of active weights) is divided equally among all passive stocks.

When zero stocks carry "Buy," all constituents receive equal weight.

Theme funds use a higher cap because concentrated portfolios with 4–5 stocks need each position to be meaningful. The cap prevents any single stock from dominating the fund while still allowing conviction to drive allocation.

Each stock's decision signal is determined by crossing its thesis status (derived from AI analysis of corporate news against four investment assumptions) with its valuation zone (derived from archetype-specific financial metrics):

Thesis status is a worst-case rollup of four assumption health scores. Each assumption is scored 0–100 by cross-referencing daily corporate news against pre-defined investment theses. ≥70 = Holding, 20–69 = At Risk, <20 = Broken. All assumptions Holding → Strong; any At Risk → Maintaining; any non-critical Broken → Weak; any critical Broken → Broken.

Net Asset Value starts at 100.00 on the inception date (T+0) and is updated daily after market close:

where:

• w_i,t−1 = weight of stock i at yesterday's close (no look-ahead bias)
• r_i,t = price return of stock i today: (P_today − P_yesterday) / P_yesterday

Weights are determined at each day's close and take effect for the following day's return calculation. This eliminates look-ahead bias.

BSE Sensex 30 Fund

Benchmarked against the S&P BSE Sensex index, rebased to 100 on T+0. The fund universe exactly equals the benchmark constituency (all 30 Sensex stocks), eliminating stock-selection bias. Any outperformance is attributable solely to the AI's conviction-weighted allocation versus the Sensex's free-float market-cap weighting.

US Tech 6 Fund

Benchmarked against an equal-weight buy-and-hold portfolio of the same six stocks (TSLA, META, AAPL, GOOGL, MU, MSFT). Each stock starts at 16.67% on T+0 and is never rebalanced — weights drift freely with price movement. This benchmark answers: does AI-driven conviction weighting beat naïve equal allocation of the same stocks?

Defence Electronics (India) — Theme Fund

Benchmarked against the BSE Sensex, rebased to 100 on the fund's own inception date. Unlike the BSE 30, this is a stock-selection test: the fund holds 4 defence companies that are not in the Sensex. Outperformance = the AI's theme-driven picks (and weighting) beat the broad Indian market.

Theme funds test a core thesis: can the AI identify secular themes and pick the companies riding them before the market recognises them?

The Defence Electronics (India) fund was constructed through the following process:

1. Theme identification: India's defence indigenisation push was identified as a structural, multi-year growth theme backed by verified government disbursement data (₹3L Cr defence production target by FY29, exports at 31× pre-2013 levels).
2. Company mapping: Small-cap companies (≤₹15,000 Cr market cap) with direct exposure to the theme were identified via industry reports, PLI beneficiary lists, and earnings data.
3. Governance gate: Companies were screened for Piotroski F-Score (>3), promoter pledge (<15%), and accounting quality.
4. Management quality: Founder-led status, capital allocation discipline, margin stability, and earnings call candour were evaluated.

The same AI thesis engine (KB generation → archetype assignment → thesis generation → daily news grading) that runs for BSE 30 and US 6 stocks monitors these companies daily. Conviction scores, decision signals, and rebalancing follow identical rules.

The portfolio is rebalanced when either of two conditions is met:

• Signal change: Any stock's decision signal changes from the previous day (e.g., Hold → Buy, or Buy → Sell).
• Drift threshold: Any stock's actual weight (after daily price drift) deviates more than 2 percentage points from its target weight.

There is no calendar-based rebalancing (e.g., monthly or quarterly). On a signal downgrade from "Buy," the stock's active position is fully sold to passive residual weight. On days without a rebalance trigger, weights drift freely with price movement.

Alpha

Arithmetic difference between cumulative portfolio return and cumulative benchmark return. Reported as cumulative since inception. CAGR is reported only after ≥90 trading days:

Sharpe Ratio (shown after ≥60 trading days)

Measures risk-adjusted return. Computed as annualized excess return over the risk-free rate, divided by annualized volatility of daily portfolio returns.

Risk-free rates: 6.5% annualized for BSE 30 (India 91-day T-bill); 5.0% for US 6 (US 3-month T-bill). Updated quarterly.

Monthly Win Rate

Percentage of complete calendar months in which the fund's return exceeded the benchmark's return. Stub months (partial first/last month) are excluded.