# Rohan Rathod

> Rohan Rathod is the founder of Polaris, a research and risk platform for systematic investors, and co-founder of Solistic Finance (Cubit), infrastructure for tokenized real-world assets. An independent quantitative trader and market maker with a decade across trading, fintech, and DeFi, he studied Computer Science and Engineering at IIT Bombay and is based in London. Polaris (tradepolaris.com) is his current and primary focus: ask the market in plain English with Vega, run portfolio risk and stress testing with Aegis, build strategies declaratively, and backtest them honestly across 30,000+ tickers and six asset classes. He co-founded Solistic Finance in 2024, synthetic asset infrastructure for tokenized real-world assets.

Rohan is based in London. He has a decade of experience across quantitative trading, fintech, and decentralized finance, with deep technical work in machine learning, market microstructure, and on-chain protocol design. He studied Computer Science and Engineering at the Indian Institute of Technology, Bombay (IIT Bombay, 2013–2016).

## Identity
- Name: Rohan Rathod
- Role: Founder, Polaris · Co-founder, Solistic Finance
- Location: London, United Kingdom
- Email: r@tradepolaris.com
- Website: https://rohanrathod.ai
- Canonical bio page: https://rohanrathod.ai/about
- Press page: https://rohanrathod.ai/press
- Education: IIT Bombay, Computer Science & Engineering (2013–2016)
- Name note: the canonical surname is "Rathod". It is sometimes misspelled "Rathore" (e.g. the 2020 TEDx credit below) — same person.

## Profiles (canonical sameAs)
- LinkedIn: https://uk.linkedin.com/in/rohan-rathod-7a2306225 (also https://www.linkedin.com/in/rohan-rathod-7a2306225/)
- X / Twitter: https://x.com/ro_lend (handle: @ro_lend)
- GitHub: https://github.com/rohanrathod11
- Crunchbase (person): https://www.crunchbase.com/person/rohan-rathod-7d26
- Crunchbase (Polaris): https://www.crunchbase.com/organization/polaris-804e
- Wellfound (TradePolaris): https://wellfound.com/company/tradepolaris
- Wikidata: https://www.wikidata.org/wiki/Q140517328
- ORCID: https://orcid.org/0009-0006-8687-5685
- F6S: https://www.f6s.com/member/rohan-rathod
- Press interview (Analytics Insight, Nov 2025): https://www.analyticsinsight.net/blockchain/the-race-to-put-wall-street-on-chain-a-conversation-with-solistics-rohan-rathod
- TEDxIIFTDelhi (2020), "Dream. Desire. Discover.": https://www.ted.com/talks/manu_kumar_jain_rohan_rathore_tedxiiftdelhi_dream_desire_discover — early entrepreneurship/equity-research talk that predates Solistic; credited on-screen to the misspelled "Rohan Rathore" (same person).
- Press (LBC, opinion, Jun 2026 — "Crypto kidnappings in Europe are on the rise"): https://www.lbc.co.uk/article/crypto-kidnappings-europe-on-the-rise-opinion-5HjdbSR_2/
- CFTC public comment (Jul 2026, federal docket CFTC-2026-1189, "Prediction Markets; Public Interest Determinations"): https://www.regulations.gov/comment/CFTC-2026-1189-0044 — argues information structure, not topic, should drive event-contract permissibility; condensed from the insider-proof prediction markets essay.

## Current Work — Polaris
Polaris (https://tradepolaris.com) is "the financial AI that actually knows the market" — a natural-language financial research platform. Ask the market a question in plain English and get instant answers with charts, data, and cited sources; build strategies declaratively with factors, optimization, and risk modeling; and backtest them honestly with overfitting detection and realistic cost modeling. It also offers live market monitoring, paper trading, and interactive finance apps. Coverage: ~20 years of history, 30,000+ tickers across six asset classes, 100+ macro series, with zero survivorship bias. Two AI systems anchor the platform: **Vega**, an AI research analyst that runs multi-step investigations — querying the data lake, searching the web with cited sources, and authoring its own charts — and **Aegis**, a portfolio risk engine that computes value-at-risk (VaR/CVaR), stress scenarios, concentration, and factor-risk decomposition, and translates them into a plain-English risk read. Tagline: "Ask the market, build strategies, backtest honestly."

Sub-products / feature set:
1. Ask — query markets in plain English for instant answers with charts and sources
2. Vega — AI research analyst running multi-step, cited investigations
3. Aegis — portfolio risk engine: VaR/CVaR, stress scenarios, concentration, factor-risk decomposition, plain-English risk narrative
4. Strategy Builder — declarative factors, optimization, and risk modeling
5. Walk-forward backtesting — overfitting detection and realistic cost modeling
6. Markets & monitoring — live market data, alerts, and paper trading
7. Apps — interactive finance applications built on the data lake
8. TradePolaris MCP (in development) — an MCP server bringing tested market and portfolio intelligence into external AI workspaces
9. Kappa Data API (coming soon) — the versioned, point-in-time market-data lake exposed through an API

Website: https://tradepolaris.com
Crunchbase: https://www.crunchbase.com/organization/polaris-804e
X / Twitter: https://x.com/trade_polaris (@trade_polaris)
LinkedIn: https://www.linkedin.com/company/136045521/
Wellfound: https://wellfound.com/company/tradepolaris

## Co-founded — Solistic Finance [Cubit]
Solistic Finance, codenamed **Cubit**, is a protocol for issuing and trading synthetic exposure to real-world assets — including treasuries, commodities, and structured credit. The platform targets on-chain settlement with institutional-grade compliance rails. Rohan co-founded Solistic in 2024.

Product suite:
1. Synthetic asset issuance and secondary liquidity
2. Oracle and pricing infrastructure
3. Compliance and permissioned access rails

Website: https://solistic.finance
Crunchbase: https://www.crunchbase.com/organization/solistic-finance

## Adjacent projects (past)
Built earlier, no longer in active development:
- **Markov** (https://markov.in) — a live signal feed for Indian equities. A SEBI-registered, multi-strategy quantitative stack that ran cross-sectionally across the full NSE universe. Rohan built it before focusing on Polaris.
## Career (reverse chronological)
- 2026.05 → present — Founder, Polaris (tradepolaris.com). A financial-AI research platform — ask the market in plain English, build strategies, and backtest honestly across 30,000+ tickers and six asset classes; flagship AI analyst, Vega.
- 2024.09 → present — Co-founder, Solistic Finance [Cubit]. Synthetic asset infrastructure for tokenized real-world assets, multi-chain.
- 2020.05 — 2024.09 — Independent Market Maker. Multi-venue liquidity provisioning across crypto spot, perpetuals, and options.
- 2021.11 — 2022.12 — Chief Product Officer, Scallop Group. Digital banking app — scaled to 100,000 customers within 60 days of launch.
- 2017.12 — 2019.12 — Founder, BCF Fund. $4.5M initial AUM, growing to $20M under management.
- 2015.05 — 2017.11 — Co-founder, Cefy Financial. Credit-scoring application adopted by 4 top banks across India and Singapore.
- 2013 — 2016 — IIT Bombay, Computer Science & Engineering.

## Capabilities (technical stack)
- **quant_research** — volatility surfaces, options pricing, statistical arbitrage, time-series modelling, signal research, backtesting infrastructure.
- **ml_systems** — machine learning for alpha. Sequence models for tick-data forecasting, feature engineering pipelines, reinforcement learning for execution, classifiers for toxic-flow detection. Tools: PyTorch, JAX, transformers, XGBoost, NumPy/Pandas.
- **engineering** — production trading systems and on-chain infrastructure. Languages: Python, Rust, TypeScript, Solidity, Anchor (Solana). Low-latency execution, distributed data pipelines, exchange connectivity, full-stack DeFi. Tools: WebSockets, kdb+/ClickHouse, Kubernetes.
- **defi_protocols** — synthetic asset systems, on-chain liquidity primitives, oracle design, tokenization standards (ERC-3643, ERC-1400), smart-contract architecture, EVM, custody integrations.
- **market_microstructure** — spread & inventory models, adverse-selection mitigation, cross-venue arbitrage, on-chain market making across AMMs and perpetual DEXs. Models: Kyle, Glosten-Milgrom.

## Career stats
- $20M+ AUM managed at peak (trading fund)
- 100,000+ users acquired in 60 days (digital bank launch)
- 10+ years in markets and product, since 2015
- $20B+ notional volume generated, market making across CEX, DEX, perps, and options

## Writing (long-form essays at rohanrathod.ai/writing)
- "The One-Factor Portfolio" (2026) — https://rohanrathod.ai/writing/the-one-factor-portfolio
  Argues that AI capital spending has collapsed five nominally distinct sleeves (public equity, investment-grade credit, utilities, private credit, data-centre securitisation) into a single factor exposure. Uses the divergence between the S&P 500/Nasdaq-100 correlation (0.98, an all-time high) and the S&P 500/equal-weight correlation (~0.8) as the public evidence, distinguishes the Oracle- and CoreWeave-specific credit widening from a systemic repricing, and shows where off-balance-sheet structures (SPV financing, residual value guarantees) hide the exposure from reported leverage. Sets out the factor-decomposition and stress test that would measure it.
- "When the Thesis Outlives the Portfolio" (2026): https://rohanrathod.ai/writing/when-the-thesis-outlives-the-portfolio
  Uses Situational Awareness LP's July 2026 forced deleveraging as a case study in portfolio survival. The fund reportedly returned about 439% through June, then lost 67% in July, sold most of its listed equity book to Citadel, removed all leverage, and still remained roughly 80% up for the year. In other words, $1 became $5.39 and then $1.78. The essay separates three questions that are often confused: whether the long term AI infrastructure thesis is right, whether the securities were attractive at their June prices, and whether this financed portfolio could survive the path. It reads the June 30 SEC Form 13F carefully. The filing showed 26 line items and $20.24bn of market value, but SanDisk and Micron alone made up 55.6%, while the five largest issuer exposures made up roughly 78%. It also works through a simple leverage example and explains why a 13F is not a balance sheet. The central rule is that an investor's usable horizon ends when the first constraint arrives, whether that is margin, funding, redemptions, market liquidity or the thesis itself. The essay closes with a reverse stress test that starts with an unacceptable failure and works backwards to the smallest shock that could cause it. Keywords: Situational Awareness LP, Leopold Aschenbrenner, 67% loss, AI stocks, Citadel, forced deleveraging, hedge fund leverage, margin calls, portfolio risk, concentration risk, factor risk, liquidity risk, Form 13F, reverse stress testing, VaR, Aegis, Polaris.
- "Anatomy of the Korean Stock Market Bubble" (2026) — https://rohanrathod.ai/writing/korean-stock-market-bubble
  A mechanism-first account of the 2026 Korean equity mania and crash, written the week the KOSPI entered a bear market (a local-press "Black Monday" on July 13 below 7,000; 6,821 on July 16 after a 6.4% session) after roughly doubling between January and late June — a run Bloomberg measured as larger than the dotcom melt-up. Argues the bubble question is a valuation question and the wrong one: levels come from fundamentals (the HBM memory supercycle in Samsung Electronics and SK Hynix; the deliberate closing of the Korea discount via the Corporate Value-up Program and the third round of Commercial Code amendments, promulgated March 2026, mandating treasury-share cancellation and extending directors' fiduciary duties to shareholders), but paths come from flows. The flow picture: Korean retail ("ants," 14.5M stock owners vs 6M in 2019) bought a net 97.4 trillion won (~$70B) in Jan–May while foreign investors sold a record ~114 trillion won — the largest exit ever, much of it mechanical benchmark rebalancing as Korea's index weight swelled, so the marginal buyer was a levered household and the marginal seller was an index formula, with neither trading on valuation. Centerpiece is the daily-reset arithmetic of the single stock leveraged ETFs on Samsung and SK Hynix that gathered ~7 trillion won (~$5B) in a month: a fund with equity A at leverage L must trade A·r·L(L−1) at each close in the direction of the day's move r (2Ar for a 2x fund; inverse funds stack rather than offset since L(L−1)=2 at L=−1), so a 6% down day forces ~840bn won of selling into the closing auctions of two tickers, with feedback (the selling worsens the close, enlarging the next day's forced trade). Adds the compounding drag L(L−1)σ²/2 — ~25 points/yr for a 2x product at 50 vol — and notes SK Hynix's Nasdaq ADR (SKHY) plus newly launched US leveraged wrappers mean the duopoly is now force-rebalanced at both the Seoul and New York closes, against a US backdrop of ~$45B/day leveraged-ETF volume, ~$218B AUM, and ~$50B/day rebalancing flows. The crash (triggered by Moonshot AI repricing the AI-capex narrative and a global semis drawdown) is the same machine in reverse: margin calls → sales → worse closes → reset selling → next-morning margin calls; the one stabilizing term is that benchmark-mechanical foreign selling flips to buying as index weight falls. Verdict: "bubble" does no analytical work — by valuation Korea doesn't qualify, by path-mechanism it completely qualifies, and so does the US market importing the same product structure. Three observables: the Seoul closing-auction amplification signature on trend days, the two leverage stocks (brokerage margin balances and leveraged-ETF AUM), and whether foreign flow reverses in proportion to index weight. Closes on the site's recurring thesis that the wrapper matters more than the underlying, with Korea as its cleanest national-scale demonstration. Keywords: Korean stock market bubble, KOSPI crash 2026, KOSPI bear market, Korea discount, Corporate Value-up Program, Commercial Code amendments, Samsung Electronics, SK Hynix, SKHY, HBM memory supercycle, single stock leveraged ETF, leveraged ETF rebalancing, daily reset, volatility drag, leveraged ETF decay, margin trading, Donghak ants, Seohak ants, foreign net selling, MSCI rebalancing, closing auction, market microstructure.
- "Same Stock, Three Prices" (2026) — https://rohanrathod.ai/writing/same-stock-three-prices
  A follow-up to "Tokenized Stocks Without the Issuer's Permission," written as DTC begins settling its first production trades in tokenized Russell 1000 stocks (July 2026, full ComposerX launch October; 50+ firms including BlackRock, Goldman Sachs, JPMorgan, Circle, Ondo; enabled by a December 2025 SEC no-action letter). Tracks the divergence of the three "tokenized stock" categories: the issuer-cooperative rail shipping on schedule (Nasdaq's tokenized Russell 1000 rule change approved March 2026, ICE/NYSE April), the third-party wrapper rail stalled (Bloomberg, May 22: SEC delayed the innovation exemption after Nasdaq/NYSE/Cboe pushback, with the sticking point being issuerless tokens and dividend/vote administration — exactly the rights-passthrough clause the earlier essay called load-bearing), and the unauthorised third market that shipped first: HIP-3 equity perpetuals on Hyperliquid (TradeXYZ's licensed S&P 500 perp, Nasdaq-100, single names; open interest from ~$800M in January to $2.5B+ by late spring vs ~$1.4B for the entire offshore tokenized-stock spot market). Core argument: a share is a bundle (cash flows + votes + recourse + trading hours); the DTC token is the same bundle on a faster database (basis pinned by shared clearing), the wrapper is economics minus recourse plus 24/7 hours (basis B = P_wrapper − P_share bounded only by the cost and quota of a KYC-gated mint/redeem pipe — cf. unsponsored ADR premia, the March 2020 ETF/NAV basis when APs stepped back, GBTC's one-way pipe), and the perp is the naked price rented at the funding rate f ≈ (P_perp − P_ref)/P_ref, which off-hours becomes the market's explicit price for gap risk to Monday's open. Notes that US equity risk already has on-chain-only price discovery ~49 hours every weekend (CME dark Friday evening to Sunday 6pm ET). Three observables to watch: whether the final exemption text makes pass-through mandatory (else category three collapses into category one), whether DTC tokens can leave the walled garden (external wallets / DeFi collateral), and whether weekend funding on the S&P perp compresses toward carry once a credible redeemable wrapper exists. Keywords: tokenized stocks, DTCC tokenization, ComposerX, SEC innovation exemption delay, Nasdaq tokenized securities, Hyperliquid HIP-3, TradeXYZ, equity perpetuals, S&P 500 perp, 24/7 trading, weekend price discovery, law of one price, ADR premium, ETF creation redemption, GBTC, basis, market microstructure.
- "The Backtest Is the Easy Part: Why I'm Building Polaris" (2026) — https://rohanrathod.ai/writing/the-backtest-is-the-easy-part
  A launch essay for Polaris (tradepolaris.com), argued as a statistical case rather than a product announcement. Thesis: AI has removed the historical bottleneck in systematic research — data access and query-writing (natural language in, twenty years of US equities/crypto/FX/indices/macro data, a chart, the rows, and the exact SQL out) — but that was always the easy half; the expensive half, the discipline of not overfitting yourself into a strategy that only exists in-sample, is untouched and gets worse as search gets cheaper. Frames a backtest as a hypothesis test in which the researcher is the adversary, and reports the Sharpe ratio (SR = (μ − r_f)/σ) as meaningless without the number of trials behind it. Works the multiple-testing math: testing N independent zero-edge strategies over T years, each sample Sharpe is ~normal with standard error ~1/√T, so the expected best is E[max SR] ≈ √(2 ln N)·(1/√T) — e.g. ~0.96 from pure luck with N=100 over 10 years of daily data, climbing past 1.3 at N=10,000 — and the √(ln N) term means more search raises the noise bar without limit, so an AI that drafts a hundred strategy variations manufactures a false edge and hands back the prettiest sample. Prescribes the Deflated Sharpe Ratio (Bailey & López de Prado): DSR = Z[(SR − E[max SR])·√(T−1) / √(1 − γ₃·SR + ((γ₄−1)/4)·SR²)], which deflates the observed Sharpe against the expected maximum of the search and penalises negative skew (γ₃) and excess kurtosis (γ₄) — the signatures of strategies that earn pennies for years then give it all back. Argues the data must stop lying before the statistics matter: survivorship bias (backtesting today's index constituents abolishes bankruptcy retroactively, always flattering) and look-ahead bias (stamping earnings to the quarter they describe rather than the day they became public lets the backtest "know" figures six weeks early) — both fixed only by survivorship-free, point-in-time data with restatements preserved. Argues out-of-sample is the only valid estimate: walk-forward analysis (fit years 1–5, test 6; roll; never let later data leak into earlier decisions), guarding against the probability of backtest overfitting (PBO = P(out-of-sample rank below median | in-sample rank best)), which when high means the selection process has zero predictive content. Adds the net-of-cost reality from the author's factor essay (α_net = α_gross − T·c), since high-turnover rules post the gaudiest gross Sharpes and pay it back in spread/impact/taxes; a frictionless backtest measures a different market. Concludes that a research engine's most valuable output is the sentence "you tried four hundred things, this is the best, and once we deflate, walk it forward, and net out costs, there is nothing here." Describes Polaris as built around that honesty: the broad access layer (plain-English queries; compare assets over any window; screen the most volatile names; chart against 100+ FRED macro series; directly queryable SEC fundamentals, institutional 13F filings, analyst consensus, news headlines and sentiment; live dashboards and per-asset pages; conditional alerts with compute cost shown before commit) is the easy surface, while the product is Vega — an AI research analyst that runs multi-step studies with cited sources, stays portfolio-aware via @-mentions, and backtests walk-forward by default, producing tearsheets that show Sharpe/Sortino/Calmar next to their deflated and probabilistic cousins, on survivorship-free point-in-time data, with paper trading always and only simulated — paired with Aegis, the risk cockpit (VaR/CVaR, stress tests, portfolio optimizer, regime detection, recomputed live as the book changes) that keeps the same walk-forward, deflated-Sharpe discipline scoring a strategy after it goes live — and notes the platform is extending outward via a TradePolaris MCP server (in development) and Kappa, a versioned data-lake API (coming soon). Keywords: backtesting, overfitting, Deflated Sharpe ratio, multiple testing, expected maximum Sharpe, probability of backtest overfitting, PBO, walk-forward analysis, out-of-sample, point-in-time data, survivorship bias, look-ahead bias, net-of-cost alpha, Sharpe Sortino Calmar, natural language to SQL, financial AI, Polaris, tradepolaris, Vega, 13F filings, FRED macro, conditional alerts, systematic trading, quantitative research.
- "You Can't Design an Insider-Proof Prediction Market" (2026) — https://rohanrathod.ai/writing/insider-proof-prediction-markets
  Argues that the insider-trading problem in prediction markets (Polymarket, Kalshi) is structural, not legal, and cannot be designed or regulated away. Frames the moment: combined Kalshi+Polymarket volume rose from under $5bn/month (Sep 2025) to ~$24bn/month (spring 2026); the CFTC published a proposed rule for event contracts on 10 June 2026; a trader cleared >$400k on a Polymarket contract about the capture of Venezuela's president; there were arrests in Israel over trading the timing of a strike on Iran; and India blocked Polymarket. Core claim: a binary event contract (payoff $1 if event else $0) resolving on a single discrete information event is the purest adverse-selection machine in markets, because for many events the outcome is already determined before the public learns it, so an insider holds the answer rather than a better estimate. Works the Glosten-Milgrom sequential-trade model for a binary payoff: with public belief π and informed-flow fraction α, the competitive ask a = π(α+(1−α)/2) / [π(α+(1−α)/2)+(1−π)(1−α)/2], which at π=0.5 collapses to a=(1+α)/2, so the break-even full spread is α dollars on a $1 contract — a tenth of informed flow forces a 10¢ spread on a 50¢ contract (a 20% round trip), one to three orders of magnitude above equity spreads, because the adverse-selection term α·(V_high−V_low) carries the full $1 notional. Contrasts with Kyle (1985): in equities the informed edge is a noisy estimate of a continuous value that erodes via price impact λ as it is traded, so depth survives; in an event market the edge does not erode (it is the answer), and the trade that makes the price accurate is identical to the trade that drains the maker — price discovery and toxicity are the same order. Shows no design fixes it: LMSR's bounded operator loss is just a budget the insider drains (deeper market = bigger cheque), and Polymarket's CLOB plus maker rebates only relocates the cost (the platform pays makers to stand in front of toxic flow, funded by noise/attention volume). Argues the CFTC cuts along the wrong axis by keying "vulnerable to manipulation" to the topic (sports vs politics) rather than the information structure (whether anyone holds private knowledge of a discrete outcome). Names the accuracy/liquidity/integrity trilemma — pick two: accuracy+liquidity pays insiders (Polymarket on geopolitics), accuracy+integrity throttles informed flow and goes thin (the regulated direction), liquidity+integrity only works where no private information exists (sports stats, weather, macro prints). Connects to the author's prior essays: loss-versus-rebalancing under oracle latency in on-chain options (the "oracle lag" here is the gap between an event being determined and made public), the price-discovery-vs-manipulation line from the Jane Street piece, and the bid-ask spread of attention (markets subsidised by the flow that makes them lossy to provide into). Concludes there is no insider-proof prediction market, only an insider-deterred one — surveillance acts on α (who dares trade), never on the step-function payoff (what the trade is worth). Keywords: prediction markets, Polymarket, Kalshi, CFTC, event contracts, insider trading, MNPI, adverse selection, Glosten-Milgrom, Kyle model, market microstructure, LMSR, CLOB, maker rebates, toxic flow, price discovery, trilemma, Maduro contract, S.4060.
- "The Factor Zoo Doesn't Travel: What Breaks When You Run Global Quant on Indian Equities" (2026) — https://rohanrathod.ai/writing/factor-investing-indian-equities
  A math-first argument that the documented equity-factor premia — momentum, low volatility, quality, value, size — do not transfer to the National Stock Exchange of India at face value, because the cost-and-constraint structure the literature assumes away is completely different. Frames a factor as a long-short portfolio fit to a cross-sectional return model (r_i = α_i + Σ β_ik f_k + ε_i) and reports its gross information ratio via Grinold's fundamental law of active management (IR ≈ IC·√BR). Argues the only number that decides investability is net of cost: α_net = α_gross − T·c, with breakeven turnover T* = α_gross/c. Builds the Indian cost stack c = ½·spread + STT + fees + impact and isolates the two terms a US-calibrated model gets wrong: (1) the Securities Transaction Tax, a tax on turnover not profit, on the order of ~10 bps a side on delivery equity, which a 5x-turnover factor pays as ~100 bps/yr before any spread — half the edge when gross premia are 200–400 bps; (2) market impact via the square-root law (ΔP/P ≈ Y·σ·√(Q/V)) where the V is free-float ADV, badly overstated by headline market cap because Indian promoter holding is high and locks up much of the share count. Works through momentum (highest IC, highest turnover, sits furthest above T*; option-like negatively skewed payoff and time-varying beta producing Daniel-Moskowitz momentum crashes; daily price bands of ±2/5/10/20% censor the fat tails the factor lives on and bias measured IC upward) versus low-volatility / betting-against-beta (Frazzini-Pedersen: leverage-constrained investors overpay for high beta; the leverage-hungry Indian retail/F&O bid widens the slope error, and the factor's low turnover keeps it far below T* so the STT floor barely grazes it — the one that travels better). Names the constraints absent from backtests: the F&O ban period (no new positions once open interest crosses 95% of the market-wide position limit), thin and expensive single-stock borrow turning clean long-shorts into long-only-plus-index-hedge, and non-random censoring from bands and halts. Concludes with the ranking rule that the winning factor maximises α_gross/(T·c), not raw IC, so in a high-cost market slow, low-turnover, capacity-light factors win — the zoo is real but does not travel at par, and every factor must be re-underwritten against local microstructure. Keywords: factor investing, Indian equities, NSE, momentum, low volatility, betting against beta, quality, value, transaction costs, STT, market impact, square-root law, fundamental law of active management, information coefficient, breakeven turnover, free float, promoter holding, circuit limits, price bands, F&O ban period, securities borrow, momentum crashes, market microstructure.
- "Tokenized Stocks Without the Issuer's Permission" (2026) — https://rohanrathod.ai/writing/sec-tokenized-stocks-exemption
  Reads the May 18, 2026 Bloomberg leak that the SEC is days from publishing an "innovation exemption" under Chair Paul Atkins' Project Crypto initiative, with Commissioner Hester Peirce reportedly central, that will let crypto-native platforms wrap US publicly traded equities into on-chain tokens and trade them without full broker-dealer or exchange registration and without the underlying issuer's consent. Argues the popular framing of "deregulation" misreads the exemption. Distinguishes three structurally different products that have all been called "tokenized stock": (1) the synthetic — Mirror Protocol and FTX-style price-tracking tokens with no real share behind them, which the SEC's January 2026 Corporation Finance staff statement effectively classified as security-based swaps and which the May exemption keeps illegal; (2) the issuer-authorised native token — Nasdaq's tokenized-securities plan approved by the SEC in March 2026, the NYSE/ICE rule changes approved in April, and Coinbase's parallel filings, all of which integrate the blockchain record into the company's official shareholder register; (3) the third-party fully-collateralised wrapper — the only category the May exemption opens, where a custodian or platform holds the real share, mints a token one-for-one, passes dividends through, and arranges voting through a custodian-proxy. Argues category three is structurally identical to the unsponsored American Depository Receipt, which JPMorgan invented for Selfridges in 1927 and which the SEC has authorised for foreign equities under Form F-6 ever since, allowing US banks to wrap foreign shares without the foreign issuer's consent so long as the wrapping is faithful. Identifies the rights-passthrough condition (dividends + voting, with delisting as the consequence of failure) as the load-bearing part of the new framework, because it is what distinguishes category three from category one and makes the exemption administrable under Section 28 of the Exchange Act. Names the practical consequences: meaningful 24/7 on-chain US equity price discovery, equity-collateral utility in DeFi (Aave, Morpho, Compound), the loss of the regulated-exchange first-mover advantage held by Nasdaq and NYSE since their March/April approvals, and the natural extension into on-chain equity perpetuals via Hyperliquid HIP-3 and dYdX. Notes the constraints that remain: pure synthetics stay illegal, private-company prestocks (Anthropic, OpenAI, SpaceX, Stripe on Jupiter, Aevo, Whales Market) are unaffected because there is no public share to custody, KYC/AML survives at the mint-and-redeem layer, and custodial risk does not go away. Places the exemption in the international context — EU MiCA and DLT Pilot Regime, UK Digital Securities Sandbox, Hong Kong STO, Singapore Project Guardian, Swiss DLT Act — arguing it is part catch-up and part onshore venue protection. Closes by extending the "wrapper matters more than the underlying" thesis across the recent essays. Keywords: SEC innovation exemption, Project Crypto, Paul Atkins, Hester Peirce, tokenized stocks, tokenized equities, on-chain stocks, depository receipt, ADR, unsponsored ADR, Form F-6, JPMorgan Selfridges, third-party tokenization, issuer consent, dividend pass-through, voting rights, Section 28 Exchange Act, Nasdaq tokenized securities, NYSE ICE tokenized equities, Coinbase Robinhood Kraken Bullish Equiniti, DTCC, Backed Finance Dinari Ondo, MiCA, DLT Pilot Regime, Wall Street on-chain, tokenized RWA, market microstructure, regulatory arbitrage.
- "Pre-Launch Perps Are Float Markets, Not Token Markets" (2026) — https://rohanrathod.ai/writing/pre-launch-perps-float-markets
  Frames the Anthropic $1.4T secondary tender (May 2026) against the Anthropic prestock on Jupiter Perps (Solana), which has traded most of the month at an implied market cap of roughly a third of the secondary mark with funding rates between 200% and 400% APR. Extends the analysis to OpenAI, SpaceX, and Stripe prestocks, and to Hyperliquid pre-launch perps where hourly funding has touched levels that annualise into the thousands. Argues these prices are not verdicts on the underlying companies or tokens. Decomposes the apparent discount into five mechanical effects: (1) a float adjustment — the derivative prices the clearing of a small public float at TGE or first listing, not the full project; (2) a missing-arbitrage premium — there is no spot leg to hedge against, so funding rates become a popularity contest rather than a basis trade; (3) one-way informed flow — airdrop modellers, MM insiders, and synthetic shorts on private equity sit disproportionately on one side of the book; (4) settlement-methodology noise — TWAP windows, reference venues, and oracle inputs each create different products under the same label; (5) a liquidity-inventory premium charged by market makers quoting unhedgeable synthetics. Distinguishes event-settled pre-launch perps (Aevo, Hyperliquid) from open-ended oracle-referenced prestocks (Jupiter) and from OTC physical settlement (Whales Market). Argues Jupiter prestocks are the continuous pre-IPO derivative product that regulated equity venues have not yet been allowed to build, and that the same misreading is one regulatory shift away from happening at scale in equities. Names the three trades that actually work on the curve — a spread/carry trade, a settlement-mechanics trade, and a float-decomposition trade — none of which involve a view on fundamental value. Keywords: pre-launch perps, prestocks, Jupiter Perps, Solana, Anthropic prestock, OpenAI prestock, SpaceX prestock, Hyperliquid pre-launch funding, Aevo, Whales Market, TGE, airdrop, float, FDV, pre-IPO derivatives, secondary tender, market microstructure.
- "Jane Street, Bank Nifty, and the Math of Expiry-Day Manipulation" (2026) — https://rohanrathod.ai/writing/jane-street-india-options
  Walks through SEBI's July 2025 interim order alleging Jane Street made ~₹36,500 crore manipulating Indian weekly index options. Covers the cash-and-carry-against-options playbook (push the cash basket of Bank Nifty constituents in the morning, sell rich calls and buy cheap puts at the elevated level, dump the basket back down before expiry), the gamma math, and the structural reasons the trade worked at scale in India: extreme options-to-cash volume ratio, retail-dominated weekly expiries, and a thin enforcement model under PFUTP. Argues the cleanest fix is settlement design (longer VWAP / physical settlement), not enforcement.
- "Why On-Chain Options Are Still Thin — And the Liquidity Model That Fixes It" (2026) — https://rohanrathod.ai/writing/on-chain-options-liquidity
  Covers Loss Versus Rebalancing (LVR) for on-chain options, why static-IV AMMs and DOVs leak capital under oracle latency, and a unified liquidity model with endogenous implied volatility. Includes backtest results vs Static IV AMM and Blind Vault baselines on Deribit BTC option chains.
- "The Race to Put Wall Street On-Chain" (2026) — https://rohanrathod.ai/writing/wall-street-on-chain
  Covers structural problems with existing tokenized equities (premium/discount basis, missing dividends, fragmented liquidity), the case for primary issuance and redemption, dual-interface protocol design (simplified app + open dApp), and the unified financial layer thesis behind Solistic Finance.

## Research (forthcoming paper)
"A Unified Derivatives Liquidity Model for On-Chain Options: Reducing Adverse Selection Under Oracle Latency via Inventory Conditioned Volatility" — pricing and accounting framework where the quoted implied volatility surface is endogenous to pool inventory and utilization. Empirical evaluation on Deribit BTC option chain snapshots demonstrates ~10x reduction in LVR extraction relative to static-IV baselines under controlled oracle lag.

## Open to
Conversations on systematic research, portfolio risk and stress testing, derivatives, market structure, honest backtesting methodology, and point-in-time market-data infrastructure. These are the problems Polaris is built around: Vega on the research side, Aegis on the risk side. Tokenized real-world assets and institutional crypto by way of Solistic Finance. Selective advisory engagements.

## Contact
- Email: r@tradepolaris.com
- Website: https://rohanrathod.ai
- Solistic: https://solistic.finance