DRAFT · Public evidence through August 18, 2026
Source status: First-party sites, regulator records, and public methodology pages were prioritized. Directory, social, vendor, and search-lead evidence is marked as secondary or unresolved.
Evidence Boundaries: Strategy, AI positioning, hiring intent, personnel, regulatory identity, performance, and deployment are separate fields. No candidate is ranked.
Source ledger: TC43-adjacent managers · global emerging managers · league tables and manager directories
Source posture: This is a public-signal map, not a performance ranking. A firm website, job description, conference appearance, regulator record, vendor case study, or directory entry describes a different fact. None, by itself, establishes live deployment, model quality, external AUM, or performance causality.
Executive Summary
- The search found no public table that combines legal identity, strategy, quant intensity, AI or GenAI evidence, personnel, and performance for small and emerging managers.
- HFR, Aurum, Preqin, With Intelligence, BarclayHedge, LSEG Lipper, and regulatory registers answer different questions. They should be joined as separate fields, not collapsed into one score.
- The new intake expands the coverage set across the UK, continental Europe, Canada, Australia, Singapore, Hong Kong and China, South Africa, New Zealand, India, Brazil, Spain, and Saudi Arabia.
- “AI manager” remains an evidence category, not a standardized industry classification. Public material ranges from quant research and machine-learning claims to named workflows, internal-capital platforms, and pre-launch products.
- The next useful artifact is a filterable, non-ranked registry keyed to legal entity, vehicle, strategy, evidence type, date, source quality, and disqualification reason.
What the public record can establish
The reviewed sources support four separate observations:
- Strategy identity. A manager may describe long/short equity, market-neutral, systematic macro, managed futures, quantitative directional, or a multi-strategy process.
- Technology positioning. A manager may describe machine learning, alternative data, language models, agents, evolutionary computing, or automation.
- Operating intent. Hiring, named technical roles, conference talks, research papers, and vendor case studies expose where a firm is investing attention.
- Legal and vehicle status. SEC, FCA, AFM, MAS, SFC, and other registers can help resolve whether an entity, adviser, or fund exists in a given jurisdiction.
These observations do not automatically connect. A quantitative strategy is not proof of machine learning. A machine-learning website is not proof of an external-capital fund. A named employee is not proof of current ownership of a system. A league-table position is not proof of an AI contribution.
Emerging and small-manager intake
The table uses evidence bands rather than ranks. “Direct” means the reviewed primary source describes an investment-management activity and a relevant technology or systematic signal. “Analogue” means the strategy is quantitative or systematic, but AI use was not explicit in the reviewed source. “Platform” means the source describes software, proprietary capital, or a pre-launch vehicle rather than an established external-capital hedge fund.
| Manager or vehicle | Region | Public strategy signal | Public AI/quant signal | Evidence boundary |
|---|---|---|---|---|
| TC43 | United States | SEC Form D and Form ADV records identify the entity and fund structures; public biography material describes a long/short investment context. | Public biography material uses machine-learning language; the reviewed sources do not disclose models, evaluation, or live deployment. | Current vehicle status, strategy detail, and implementation remain open checks. |
| Machina Capital | France / Singapore | Quantitative equities and equity derivatives, with public systematic-investment positioning. | The reviewed public material includes machine-learning and data-science recruiting signals; an Amundi announcement adds a strategy-level ML reference. | Model family, evaluation, production architecture, and performance attribution are not public in the reviewed set. |
| Castle Ridge Asset Management | Canada / United States | Market-neutral long/short equity is described on the firm site. | The firm describes W.A.L.L.A.C.E. using machine learning and evolutionary computing and publishes a technical team page. | The AI description is firm-reported; current fund activity and independent operating evidence require separate checks. |
| Seldon Capital | United States | Official material describes a systematic, cross-asset, long/short process. | LinkedIn positioning references machine learning; the reviewed official pages do not disclose a GenAI architecture or detailed model workflow. | The difference between official-site language and social-profile language remains unresolved. |
| Machine Capital | Netherlands | The firm describes systematic funds and machine-learning investment strategies. | Machine learning and automation are explicit in the public positioning. | Current fund status, AFM record, model details, and independent performance evidence require verification. |
| Meridian & Saturn Capital | Singapore | Public material describes systematic long/short equity across China and global markets. | The firm describes ML signals, deep learning, and dedicated infrastructure. | Claims are firm-reported; legal-entity dates, personnel identity, vehicle status, and the relationship between infrastructure and live portfolios need reconciliation. |
| Bayswater Technologies | United Kingdom | The site describes a market- and factor-neutral long/short equity strategy. | It describes research agents, a temporal-epistemic world model, and unstructured data, with a human portfolio-manager decision point. | Fund launch, external capital, regulator status, live deployment, and model evaluation are not established by the reviewed page. |
| One Eleven Capital | France | Quantitative equities and listed derivatives, with mid-frequency strategies and a stated institutional-SMA route. | The site describes quantitative and machine-learning techniques. | Launch and operating statements are firm-reported; the reviewed page does not provide audited results or model details. |
| The Maritime Fund | Canada | Long/short equities, commodities, global macro, and a machine-learning group are described. | The ML group is said to develop quantitative strategies and AI tools for risk metrics, screening, and idea generation. | The reviewed page does not establish live ML performance or the boundary between investment models and internal tools. |
| JCube Capital Partners | Singapore | The firm describes systematic fund management, accredited/institutional clients, and Singapore-domiciled fund vehicles. | Its first-party page names statistical models, empirical research, machine-learning signals, alternative data, and large-language-model outputs in the public research framing. | The MAS licence statement should be checked against the MAS directory; fund names, model architecture, training data, and independent outcomes are not disclosed. |
| Aggregate Asset Management | Singapore | First-party material describes a global-equities/value-fund process and names the people involved in developing it. | Aggregate says its proprietary “Deep Deep” machine-learning model is being expanded after a multi-year development period; an earlier first-party article describes machine learning for stock selection. | Firm-reported evidence; no model weights, evaluation protocol, training corpus, permissions, or independent results. |
| ChengQi Funds | Hong Kong / China | Quantitative fund management and private-fund activity are described. | The site references big-data processing, machine learning, and AI across research, model development, risk control, and trading. | Current model evidence, regulatory records, and independently confirmed scale remain open checks. |
| High-Flyer Quant / DeepSeek | Mainland China | High-Flyer’s current English quant page describes quantitative investing, regulated private-fund subsidiaries in mainland China, and a Hong Kong Type 9 entity. | The firm says it began exploring fully automated trading with machine learning in 2008, used a deep-learning trade in 2016, built the Fire-Flyer deep-learning platform, maintains more than 10PB of data and thousands of sources, and uses neural networks and NLP. These are firm statements, not independent technical audits. | The page does not establish current trading-model inventory, whether DeepSeek research is used in trading, current personnel allocation, data rights, or audited investment outcomes. |
| JoinQuant Investment | Mainland China | The Chinese first-party site identifies JoinQuant Investment as a private-fund manager registered with AMAC and says it was established in 2017. | Its public AI-research surface describes an “AI autonomous-driven” investment-research direction, a factor/model/optimization/decision structure, and a team that includes people with ICML and ICLR publications. The page does not specify which models or papers are used in live portfolios. | First-party positioning and registration number; no current vehicle list, model cards, training data, production authority, or independent performance evidence was recovered. |
| Ubiquant Investment | Mainland China | The official competition page describes the UBIQUANT CHALLENGE, an AI reasoning challenge, and full-time/internship routes. The UbiquantAI Hugging Face organization links to the firm website and lists public AI model/paper activity. | Clocktower’s 2024 China Quant Deep Dive names Data, AI, and Waterdrop labs for data, algorithms, and trade execution. The public Fleming-VL-8B model card describes medical multimodal reasoning and explicitly lists UbiquantAI releases; the reviewed material does not connect those models to finance. | Public AI-research, competition, and recruiting surfaces are now stronger than the prior secondary-only record, but current lab leadership, filled roles, investment model inventory, data rights, live authority, and outcomes remain undisclosed. |
| Wright Research | India | First-party pages describe a quant PMS, robo-advisory, factor strategies, and regime-aware portfolio products; the India Opportunity episode identifies Sonam Srivastava as founder and portfolio manager. | Wright publicly describes AI/ML forecasting and allocation. Its product surface names CTO Vinod Reddy Kotha and ML adviser Dr Miquel Alonso. | These are first-party product and biography claims plus podcast metadata. The reviewed set does not establish model architecture, training data, permissions, live authority, or independently audited performance. |
| Saudi Exchange Algorithm-Enhanced Trading Fund | Saudi Arabia / United Kingdom | The Arabic prospectus documents a systematic Saudi-equity vehicle and separates the Saudi Fransi Capital fund-manager role from Winton’s investment-adviser role. | The prospectus names algorithmic return and risk prediction, portfolio construction, cost control, data processing, target-weight assignment, and program updates; it assigns algorithm development and operation to Winton. | Regulatory prospectus terms establish a documented design and responsibility map, not live scale, realized performance, model weights, training data, or post-prospectus implementation. See the capture note. |
| Golden Hen Asset Management | Hong Kong | Quantitative futures trading and Hong Kong Type 4 and Type 9 licences are described. | The firm references generative AI and collaboration with HKU’s FinTech Academy. | Verify current permissions in the SFC register; model and evaluation evidence are not public in the reviewed page. |
| Avangard Investments | Australia | A systematic Australian-equity fund is described as beginning July 1, 2026. | The strategy page names A.L.F.R.E.D., an adaptive-learning ranking system for equities and derivatives, with traceable data/rule outputs. | The page says investment-team review remains in the loop; it does not provide implementation details, data rights, independent results, or evidence of generative AI use. |
| STANLIB Systematic Solutions | South Africa | A systematic-investment division is identified through a named quantitative-research leader. | Chetan Ramlall’s interview discusses machine learning, alternative data, satellite/drone imagery, automation, and risk quantification. | Publisher interview and guest account; current model inventory, permissions, and independent outcomes remain open. |
| NMRQL Research | South Africa | The dated episode identifies NMRQL as an investment-management firm making adaptive, testable trading decisions. | Chief Scientist Stuart Reid discusses machine learning/deep learning for non-stationary financial time series. | Historical podcast evidence; current legal status, personnel, models, permissions, and live results require re-checking. |
| Index Solutions | South Africa | Quantitative investment management and automated portfolio management are described. | The firm says it combines statistical techniques with in-house machine learning. | Hedge-fund structure, long/short exposure, external AUM, and audited performance are not established by the reviewed source. |
| QuantOptimus | Canada | Statistical-arbitrage signals and automated hedge baskets are described. | The platform describes LLM-powered reasoning agents and risk-manager agents. | The source describes an analytics software company; it is not evidence of an external-capital hedge fund or live third-party deployment. |
| 168 Capital Management | Hong Kong | The site describes proprietary-capital research rather than client-fund management. | It references AI agents, AI risk controls, and proprietary model R&D. | The site explicitly separates the activity from accepting external funds; treat it as an internal-capital technology signal. |
| Phantom Capital | South Africa | A quantitative futures system is described as running on personal capital. | ML filters, retraining, and loss controls are described. | The site says it is not a registered investment adviser; published performance figures are self-reported. |
| Binomial Orion | United States / emerging | The public site describes a long/short equity product. | Sentry, specialist models, and 700-plus signals are described. | The reviewed site still labels the product “pre-launch”; no regulator record was located in this pass. |
Global discovery leads requiring another pass
The search also surfaced Bateson Asset Management, Noax Capital, Plurimi AI Long/Short Equity, Infinitus Capital, Gan IQ, RQI Investors, UNKAI, Pryon Capital, High-Flyer, Ubiquant, Storm Capital, Protea Capital, RegentQuant, Salt Funds Management, Sparkline, AXQ, CapitalsAI, Nujum, XYZ Capital, PharVision Capital, Signum Investments, Move37 Capital, Marduci Capital, Runtime Fund, Voleon, and High-Flyer-related sources.
These names are not promoted into the evidence table merely because a search result, directory, or social profile exists. The next pass must resolve the legal entity, vehicle, current status, strategy, and source provenance for each. The global discovery ledger records the inclusion reason and the missing check for each lead.
What “league table” means in this market
There is no single public league table that measures the requested dimensions together. The public and commercial layers describe different objects:
| Layer | What it can describe | What it cannot establish |
|---|---|---|
| HFR strategy classifications and HFR database | Strategy buckets, fund descriptions, assets, liquidity, and reported monthly performance in a commercial database. | GenAI depth, model ownership, or whether a strategy’s public label matches its internal process. |
| Preqin League Tables | Dated performance tables using defined eligibility rules and strategy filters. | A complete manager census; Preqin says public tables use public and participant-reported information and are not independently verified. |
| With Intelligence indices | Strategy and regional benchmarks, index constituents, and methodology. | A complete universe; index selection, AUM screens, turnover rules, and survivorship adjustments shape the result. |
| Aurum Strategy Engine | Strategy-normalized performance, AUM, flows, and qualitative manager data in a commercial engine. | Public-complete manager records or a public AI classification. |
| BarclayHedge indices | Hedge-fund and managed-futures index returns and selected universes. | An AI ranking or a census of small managers; inclusion and reporting rules matter. |
| LSEG Lipper and Morningstar categories | Fund data and category structures. | Proof that a category label reflects the manager’s internal technology. |
| SEC IAPD, FCA register, AFM registers, MAS directory, and SFC register | Entity identity, permissions, filings, and jurisdictional status. | Performance, AI implementation, current personnel ownership, or investment quality. |
The right output is therefore a registry with separate columns for strategy, AI signal, personnel, regulatory status, performance source, date, and disqualification reason. A single composite score would hide the exact evidence boundaries that make this research useful.
Recommended registry fields
Use one record per manager and vehicle, with explicit dates:
legal_name · public_brand · fund_or_vehicle · jurisdiction · regulatory_record · manager_status · fund_status · strategy_primary_source · strategy_normalized_hfr · quantitative_systematic_evidence · ai_ml_claim_type · ai_evidence_tier · ai_source_date · personnel_source · performance_source · performance_period · aum_source · aum_date · selection_bias_note · disqualification_reason · open_question
The AI evidence field should remain categorical: none located, public positioning, hiring intent, named workflow, firm-owned artifact, production claim, or independently evaluated. A league-table position must not upgrade that field.
What to ingest next
The next passes should run in this order:
- Resolve every U.S. candidate through SEC IAPD, Form D, and Form ADV, including aliases and fund vehicles.
- Resolve foreign managers through the relevant regulator and corporate register before treating licensing or live-fund claims as established.
- Capture personnel only from current first-party pages, regulator filings, conference rosters, papers, or direct public artifacts. LinkedIn and The Org are discovery inputs, not standalone proof.
- Add podcast, keynote, GitHub, job, and vendor surfaces only after entity resolution. Preserve the source date and the exact claim type.
- Keep a separate performance table keyed to metric, period, eligibility, reporting source, and selection bias. Do not merge it with the AI evidence map.
Sources and method
The underlying ledgers contain the retrieved URLs, dates, confidence labels, inclusion and exclusion notes, and follow-up checks: TC43-adjacent managers, small quant managers, global emerging managers, and league tables and manager directories. The pass prioritized first-party websites, regulator records, technical or product pages, and official methodology documents; directories, aggregators, press releases, and social profiles remain marked as secondary or discovery evidence.
The research does not infer that any named firm is ahead, behind, better, or worse. It records where public evidence exists, what the source actually says, and what remains unverified.
State of AI | August 2026