A Portfolio Construction Framework helps investors by providing a structured approach to building a diversified investment portfolio. Based on assumptions that each asset is not mutually perfectly correlated, it aims to optimize risk and return, ensuring alignment with individual goals and portfolio constraints. In addition to traditional approaches such as Modern Portfolio Theory, we incorporate machine learning techniques to enhance predictive accuracy of asset behavior, enabling more adaptive and data-driven portfolio decisions. This combination supports better risk-adjusted returns and more informed, forward-looking investment strategies to better support investment decision and benefit investors.

Our Portfolio Construction Framework is a disciplined and adaptive approach to building high-conviction investment portfolios. It integrates traditional financial theory—such as risk-return optimization and diversification—with advanced machine learning models that enhance predictive accuracy and market responsiveness. Our portfolio is thoughtfully divided into three core strategy pillars: - Event Driven: Focused on catalysts such as mergers, restructurings, and special situations, leveraging deep fundamental and risk arbitrage expertise. - Equity Long Only: High-conviction, research-driven equity positions targeting structural winners across global markets (incl. Engagement Strategy) - Quantitative & Other Hedge Fund Strategies: AI-enhanced, rules-based models applied to alpha extraction, asset allocation, and risk management. This multi-strategy framework allows Abundia to dynamically adjust exposures, optimize risk-adjusted returns, and align capital deployment with both macro and micro-level signals. By combining human judgment with AI-enhanced systems, we aim to build portfolios that are resilient, opportunistic, and strategically positioned for long-term performance.

We believe that engagement is a powerful and underutilized tool to unlock hidden value in public equities — not only for shareholders, but for the broader economy. Our approach focuses on identifying companies with mispriced potential due to structural inefficiencies, poor governance, or misaligned capital allocation. Through constructive engagement — whether via shareholder dialogue, strategic suggestions, or capital structure advocacy — we aim to catalyze changes that can lead to: - Improved operational efficiency - More disciplined capital deployment - Stronger governance and transparency - Better capital market signaling This not only creates alpha for our investors, but also promotes healthier capital markets and more efficient allocation of resources across the economy. In many cases, our engagement helps businesses reinvest more productively, innovate, or return capital where growth is limited, all of which contribute to better productivity and capital circulation. In this sense, engagement strategy is not just a return-enhancing mechanism, but a way to create positive economic externalities, aligning long-term shareholder value with real-economy impact.

Building a Portfolio Construction Framework begins with defining clear investment objectives and assessing the investor’s risk profile. Next, suitable asset classes are selected and strategically allocated to reflect those goals. The process includes setting diversification guidelines, establishing monitoring tools, and determining rebalancing rules. Continuous evaluation and refinement help the framework remain effective over time.

Risk management is a foundational element of the framework, aiming to protect capital while enabling growth. It involves identifying potential risks—such as market, credit, or liquidity risk—and implementing strategies to mitigate them through diversification, position sizing, and scenario analysis. A robust risk management approach ensures that the portfolio stays aligned with its intended risk-return profile across market cycles.

Deep learning models like LSTM (Long Short-Term Memory networks) are particularly powerful for time-series forecasting, which is central to investment decision-making. In financial markets, asset prices are driven by sequences of events — macro indicators, earnings revisions, order flow, sentiment, etc. Traditional models often struggle with non-linear dependencies, long memory, or noisy signals. LSTM networks are designed to address these exact challenges. We train an LSTM model using historical data, feeding in these sequences. Unlike simple regressions, LSTM can learn patterns across time. Once trained, the model provides a forward prediction — i.e., a probability-adjusted expectation of return over the next forward looking time period.

AI is integrated throughout Abundia’s investment workflow, from idea generation to portfolio optimization. Machine learning models help identify regime shifts, pattern correlations, and behavioral anomalies that may not be visible through traditional analysis. These insights complement human judgment, allowing the team to act decisively yet objectively based on data-driven probabilities.

Engagement typically follows quantitative or fundamental signals indicating governance or valuation gaps. Once identified, we pursue constructive dialogue focused on operational efficiency and capital discipline.

While many quant funds rely on static rule-based models, Abundia employs adaptive AI systems that continuously learn from new market data. Our models are designed to evolve as correlations change, ensuring our strategies remain resilient in volatile and non-stationary environments.

It acts as a diversification layer, capturing non-directional returns that are uncorrelated with event-driven or equity exposures, improving portfolio stability and risk-adjusted performance.

We view AI as an augmentation tool, not a replacement. Investment committees use model outputs as probabilistic guidance, integrating them with qualitative research, macroeconomic context, and experience-based intuition to make balanced, conviction-driven decisions.

Abundia’s philosophy centers on adaptability. By combining Event-Driven, Equity Long-Only, and Quantitative strategies under one unified risk framework, we aim to capture diverse alpha sources while maintaining disciplined exposure control and consistent risk-adjusted returns.

We focus on structural inefficiencies that arise from behavioral biases, regulatory asymmetries, or capital misallocations. Through data analytics and cross-market modeling, we identify dislocations early, often before they become visible to traditional investors.

Data integrity is foundational. We combine traditional financial data with alternative datasets, such as transaction flows, sentiment indicators, and supply-chain metrics, all subjected to rigorous cleaning, normalization, and bias testing before use in model training.

We prioritize interpretability through model validation tools and post-training diagnostics. Every prediction or signal is accompanied by an attribution layer that explains key contributing variables, ensuring transparency and governance in decision-making.

Our engagement work often stems from signals identified by our quantitative framework. When data highlights undervalued firms constrained by governance inefficiencies, we apply engagement to unlock value through direct, constructive dialogue with management teams.

Cross-strategy exposure is continuously monitored through a dynamic covariance model. Scenario testing and stress simulations ensure that even under extreme tail events, portfolio drawdowns remain within pre-defined risk limits.

Deep learning models are used not only for forecasting returns but also for predicting volatility clusters, liquidity shocks, and systemic contagion risks. This predictive layer enhances our ability to anticipate and mitigate downside scenarios.

We focus on structural positioning, aligning capital with long-term secular trends, while using tactical AI models to manage short-term timing risk. The goal is to participate in compounding themes without being disrupted by temporary volatility.

Adaptive investing refers to the continuous recalibration of models, assumptions, and exposures based on real-time data. Our systems learn from evolving market behavior, allowing strategies to remain effective even as old relationships break down.

Through active risk budgeting, hedging overlays, and multi-layered diversification. AI-based early-warning indicators help identify volatility spikes or macro regime shifts, enabling timely risk reduction without compromising long-term positioning.

Macroeconomic variables form an essential context for both model training and discretionary interpretation. Our AI systems incorporate macro time-series, such as interest rate curves, credit spreads, and inflation expectations, to refine probability outcomes.

Each model undergoes rolling backtests, live simulations, and decay analysis to assess predictive persistence. Models that lose statistical significance are retrained or replaced, ensuring the research stack remains robust and current.

Abundia continuously monitors the evolution of global markets, however, we do not invest in digital assets or cryptocurrencies under any circumstances. This is a firm policy designed to avoid speculative markets that lack transparency, regulatory consistency, and sufficient liquidity. Our framework focuses strictly on regulated, institutional-grade asset classes, ensuring adherence to our risk standards and investor protection principles.

We maintain an open, interdisciplinary research environment that merges data science with fundamental reasoning. Analysts and engineers collaborate closely to bridge the gap between quantitative modeling and real-world market interpretation.

Scalability is addressed at both capital and infrastructure levels. Our AI infrastructure allows parallelized computation and multi-market deployment, while strategy capacity is managed to prevent alpha dilution.

We envision AI evolving from a forecasting tool into a strategic co-decision partner. As models grow more context-aware and interpretable, they will enhance, not replace, human insight, ultimately helping investors allocate capital more intelligently across an increasingly complex global landscape.

Abundia combines machine learning with fundamental research, allowing models to adapt to changing market regimes. Unlike static quant funds, our approach blends data-driven precision with human judgment, creating strategies that are both systematic and fundamentally informed.

AI models help detect early signals of corporate actions, from sentiment shifts to abnormal volume, improving our ability to position ahead of key catalysts.

Our models continuously recalibrate based on market-implied probabilities, antitrust developments, and news sentiment, enabling a faster, data-driven response to changing deal conditions.