BDO Technologies Inc., 6588 Nelson Avenue, Burnaby, BC (2026)

02/07/2026

AI-SOR Framework

The AI-enhanced Stimulus–Organism–Response (AI-SOR) framework extends the classical S-O-R model by formalizing the Organism component as an AI-based cognitive system that captures behavioral decision-making under uncertainty. In this framework, external stimuli such as market prices, volatility shocks, and unstructured information constitute the stimulus layer. These inputs are processed through an AI-driven Organism layer that performs perception, preference learning, and belief updating.

From a behavioral finance perspective, the Organism layer operationalizes latent psychological constructs including loss aversion, mental accounting, sentiment, and time-varying risk aversion. By explicitly modeling bounded rationality, AI-SOR explains asymmetric reactions to gains and losses and the amplification of behavior during periods of market stress. The response layer generates constrained, explainable decision signals—such as portfolio adjustment recommendations or risk warnings—rather than autonomous trading actions. In robo-advisory applications, AI-SOR serves as a behavioral interface that converts investor-specific sentiment and belief updates into structured, confidence-weighted views compatible with portfolio construction frameworks such as the Black–Litterman model, thereby preserving diversification, risk discipline, and fiduciary consistency.

01/30/2026

Wealthfront completed its initial public offering (IPO) on December 12, 2025, listing on the NASDAQ Global Select Market. The company priced its IPO at US$14 per share, raising approximately USD 480–500 million in gross proceeds. Following the offering, Wealthfront was valued at approximately USD 2.0–2.6 billion, depending on post-IPO market pricing.

Founded in 2008, Wealthfront is one of the earliest and most prominent robo-advisor platforms in the United States. Prior to its IPO, Wealthfront served over one million clients and managed roughly USD 85–90 billion in assets under management (AUM).

Wealthfront’s IPO is widely viewed as a milestone for the robo-advisor and digital wealth management industry. It became the first large, independent robo-advisor to successfully list on a major U.S. exchange, setting a valuation reference point for the sector. The IPO demonstrates that automated wealth management can achieve sufficient scale, regulatory maturity, and revenue stability to access public markets.

It also highlights a regulatory-friendly model where automation and AI support portfolio construction, risk management, and client experience, while remaining compliant with fiduciary and suitability requirements.

01/17/2026

AI hallucination risk is not inherently unavoidable; however, if inadequately governed, it poses a systemic threat to portfolio performance, regulatory compliance, and long-term credibility of robo-advisory platforms.
AI should be strictly confined to information processing and perspective enhancement, while all investment decisions are executed within a rule-based, transparent, and auditable investment framework.

💡 Decision Separation Mechanism
➡️ AI models are limited to generating sentiment factors, thematic indicators, or view adjustments
➡️ Final asset allocation and portfolio construction are performed using established financial models (e.g., mean-variance optimization, risk budgeting, or the Black–Litterman framework)
➡️ AI outputs cannot override predefined risk constraints, position limits, or drawdown controls

📋 Cross-Validation and Consistency Checks
➡️ Multiple models and data sources are used to cross-validate critical signals
➡️ Signals that lack consistency across time or logical dimensions are down-weighted or excluded
➡️ Minimum confidence thresholds are enforced; signals below threshold levels are not incorporated into the investment process

🗣️ Explainability and Auditability Requirements
All AI-generated signals must satisfy the following criteria:
➡️ Clearly identified data sources and generation pathways
➡️ Traceable historical performance, including documented failure scenarios
➡️ Explicit explanation of how the signal informs investment views rather than directly triggering trades
Any AI output that fails to meet explainability standards is deemed invalid for portfolio implementation.

🧑‍💼 Human-in-the-Loop Oversight
➡️ In periods of elevated market stress or heightened model uncertainty, the system activates manual review protocols
➡️ Investment committees or risk managers retain full authority to pause, adjust, or reject AI-generated signals
➡️ This ensures that AI systems remain fully subject to human governance and control

By constraining AI within a rule-based, explainable, and risk-controlled investment architecture, the system effectively harnesses AI’s analytical strengths while mitigating the adverse effects of hallucinated outputs.

01/10/2026

AI Hallucination in Robo-Advisors

AI hallucination risk refers to the possibility that AI models generate outputs that appear coherent and plausible but lack sufficient factual, data-driven, or causal support, particularly in environments characterized by incomplete, ambiguous, or noisy information.

In a robo-advisory context, this risk primarily manifests as incorrect investment signal generation, rather than traditional market-driven price volatility.

Unlike market risk, AI hallucination risk is classified as a form of model risk, arising from model architecture, training data limitations, inference mechanisms, and uncertainty in probabilistic reasoning, rather than from the underlying assets themselves.

AI hallucination risk may emerge at multiple stages of the investment process, including:
1️⃣ Unstructured Data Analysis
When processing news articles, social media content, earnings reports, or analyst commentary, AI models may misinterpret semantic nuances, rely on incomplete information, or inadvertently incorporate misinformation, resulting in distorted sentiment or event signals.
2️⃣ Investment Logic Inference
AI models may incorrectly infer causal relationships from mere correlations, producing investment rationales that lack empirical, statistical, or economic foundations.
3️⃣ Low-Frequency or Extreme Market Conditions
During black-swan events, geopolitical shocks, or periods of constrained liquidity, historical reference data may be limited. In such circumstances, models are forced to extrapolate, significantly increasing the probability of hallucinated outputs.

Potential Impact of Unmanaged Hallucination Risk
If left inadequately controlled, AI hallucination risk may result in:
✔️ Generation of erroneous investment signals, leading to irrational portfolio rebalancing or excessive turnover
✔️Distortion of risk assessment metrics and weakening of portfolio risk controls
✔️Reduced explainability of investment decisions, increasing compliance and audit challenges
✔️Erosion of investor trust and long-term client retention
✔️Potential regulatory concerns related to fiduciary duty and suitability obligations
These kinds of impact may lead to disastrous losses. Therefore, AI Hallucination should be taken into account in the risk management framework. We’ll discuss about how to implement it next time.

01/01/2026

🧠 Black-Litterman × NLP × Robo-Advisor

Natural Language Processing (NLP) can be integrated into an AI-driven robo-advisor by transforming unstructured textual information into quantitative inputs that are systematically incorporated into the Black-Litterman (BL) framework.

1️⃣ NLP is applied to unstructured data sources such as financial news, corporate disclosures, government policy or social media content. These textual data contain forward-looking information about market sentiment, expectations, and uncertainty, which are not captured by traditional price-based indicators.

2️⃣ NLP techniques—such as sentiment analysis, topic modeling, and semantic embedding—convert textual information into numerical signals. These signals include sentiment scores, topic probabilities, and measures of uncertainty or disagreement. The outputs are structured features rather than direct return forecasts.

3️⃣ The quantified NLP outputs are translated into investor views within the Black-Litterman framework. Specifically, NLP-derived signals form the view vector (Q), representing directional expectations on asset or sector returns. At the same time, the confidence in these views is captured by the uncertainty matrix (Ω), where higher textual disagreement or noise implies lower confidence in the view.

4️⃣ The Black-Litterman model combines the market equilibrium returns (the prior) with the NLP-based views to generate posterior expected returns. This Bayesian approach ensures that AI-generated signals adjust, rather than replace, the market-implied equilibrium, thereby reducing the risk of overreacting to noisy sentiment data.

5️⃣ Finally, the posterior expected returns are used in portfolio optimization subject to risk constraints such as volatility limits, drawdown control, and regulatory requirements. NLP signals may also influence risk budgeting and rebalancing frequency, for example by reducing equity exposure during periods of elevated negative sentiment or uncertainty.

In summary, integrating NLP into the Black-Litterman framework allows a robo-advisor to systematically incorporate qualitative information into disciplined, explainable, and risk-controlled portfolio decisions without abandoning traditional asset pricing foundations.

12/29/2025

Traditional robo-advisors rely on structured data,like historical rate of return,volatility,correlation coefficient or financial indicators,etc. The common problem refers to this kind of data is lagging and consensus-driven,which may provide false signals.

Unstructured data refers to information without a fixed tabular format, such as text, images, audio, or video, which can’t be directly processed by traditional databases,for example,news & financial reports,social media sentiment,earnings call transcripts and search & behavioral data.

Robo-advisors transform unstructured data into quantitative signals using AI, enhancing dynamic allocation and forward-looking risk management.Typical applications of unstructured data include:

1️⃣ Emotion-Driven Dynamic Allocation
📉 Market panic → Increase defensive assets (bonds, cash, low-volatility ETFs)
📈 Market overheating → Automatic de-risking and profit taking
Based on sentiment extracted from news, social media, and market narratives

2️⃣ Risk Early-Warning System
Unstructured data enables early identification of:
🏦 Policy and regulatory risks
🏢 Credit events and corporate distress
🦢 Rising probability of black-swan events
Signals often emerge before price movements

3️⃣ Personalized Robo-Advisory
🗣️ Analyze client behavioral text(such as Q&A records,investment preference expressions) by NLP(Natural Language Processing)
🙎 Dynamically adjust risk profiles using behavioral finance insights

However,there are still plenty of challenges in using unstructured data :
🩺 High level of data noise and difficulty in distinguishing reliable information from misinformation
🎙️ Limited model interpretablity and explainablility
🔒 Privacy protection and regulatory compliance issues

12/20/2025

📈 Use of VIX in AI investment Models

The VIX (Volatility Index) measures the market’s expectation of 30-day forward volatility implied from S&P 500 index option prices,calculated by BSM model. It is commonly interpreted as a gauge of systematic risk and investor risk aversion,and widely used in real investment scenarios.

Now the VIX is taken into account in AI-based investment,which is one of the variables in our model. The VIX is incorporated as an exogenous risk factor rather than a return predictor. Models typically transform the VIX into derived features,such as:VIX level,Rate of change (ΔVIX),VIX momentum (short vs long average),VIX percentile (relative to history) or VIX regime labels (low / medium / high volatility). These features enhance the model’s ability to understand current market risk conditions, not just price direction.

AI models often classify markets into volatility regimes: Low Volatility(25,Defensive / tail-risk).Once the regime is detected, the AI switches strategies automatically. For examples:

Dynamic Asset Allocation
The VIX enables AI systems to implement dynamic asset allocation. While implied volatility increases, models typically:
✔️ Reduce exposure to risky assets,e.g.tech sector stocks,commodities;
✔️Increase allocations to defensive or low-beta assets,.e.g.utility sector stocks,fixed income assets;
✔️ Adjust leverage to maintain a target volatility or risk budget.
This improves risk-adjusted returns over the investment cycle.

Providing Trading Signals
(1) Entry / Exit Filters
AI models may:
✔️ Block buy signals when VIX is extremely high;
✔️Delay exits in low-VIX trending markets;
✔️Identify panic-selling opportunities when VIX spikes rapidly.

(2) Mean Reversion & Tail Risk
Since VIX is mean-reverting, AI systems:
✔️Predict volatility normalization;
✔️ Trade volatility products;
✔️Hedge portfolios when VIX deviates from historical norms.

Moreover, the VIX can be used in reinforcement learning as a state vector.The agent learns optimal actions by maximizing expected reward subject to volatility-dependent penalties, such as drawdowns during high-VIX environments.

12/13/2025

I took part in the event of VTJTalks: Grading Vancouver's Innovation Economy in 2025 and was inspired by the conversations and networking! The founders and institutional investors shared their experience and point of views on Vancouver’s tech and innovation scene,and their prospect of trends and challenges on funding,talent and growth,etc.

12/06/2025

The AI Investment Workflow

Modern investment teams use AI as a multistage workflow that links data engineering, model development, explainability, and deployment.

-Data: Structured and unstructured data with hidden information from market relationships to textual sentiment.
-Feature Modeling: Working out domain-specific features such as basis, skewness, and network centrality to enhance predictive modeling.
-Modeling: Building models to handle non-linearity, sequence, optimization, and decision-making under uncertainty.
-Explainability: Complying with transparency, validation, and interpretability by using SHAP (SHapley Additive exPlanations) values, visual networks, and governance frameworks.
-Decision-making: Using AI to provide information for allocation, forecasting, and strategic communication.
-Deployment: Responsible implementation and ongoing monitoring in production environments.

Together, AI workflow unfolds from data collection to deployment, forming a connected system of tools and techniques for modern investment teams.

References:
Joseph Simonian,PhD.2025.“AI in Asset Management: Tools, Applications, and Frontiers”.

11/29/2025

On Nov 24,US President Donald Trump signed an Executive Order launching “the Genesis Mission”, a new national effort to use artificial intelligence (AI) to transform how scientific research is conducted and accelerate the speed of scientific discovery.

-The Genesis Mission charges the Secretary of Energy with leveraging National Laboratories to unite America’s brightest minds, most powerful computers, and vast scientific data into one cooperative system for research.

-The Order directs the Department of Energy to establish a closed-loop AI experimentation platform, integrating US supercomputers and unique data assets to generate foundation models and support robotic laboratories.

-The Order instructs the Assistant to the President for Science and Technology (APST) to coordinate the national initiative and the integration of data and infrastructure from across the Federal government.

-The Secretary of Energy, APST, and the Special Advisor for AI & Crypto will collaborate with academia and private-sector innovators to support and enhance the Genesis Mission.

The Order is in accord with the AI policy blueprint issued by President Trump in July this year. The blueprint encourages agencies such as the Department of Energy to invest in 'automated cloud computing laboratories',involving engineering, materials science, chemistry, biology and neuroscience,and expand AI research and training within national laboratories.

Previously, Trump had appealed to legislators via social media to enact federal AI regulatory standards, suggesting that they should include them in forthcoming Defense Appropriation Act or push them forward as standalone act.

11/23/2025

I was honored to attend the business event presented by Global Financial Impact. Several senior executives from GFI gave their point of view on finanical market and expectation of the trends,especially the combination of finance and technology.The conversations with these professionals gave me some guideline and experience to be a better fintech.

11/22/2025

With the rapid development of financial technology, robo-advisors are becoming the focus of the wealth-management sector, changing the way how people manage their fortune. Generally, robo-advisors use artificial intelligence and big data technologies, based on modern portfolio theory, to provide investors automated and customized financial advice and asset allocation portfolios.

Traditional wealth-management services typically rely on human, like financial advisors. Advisors usually communicate with clients in person, understand their financial situation, investment objectives, and risk tolerance, and then design investment plans based on those information. Although this approach offers a degree of customization, it is relatively inefficient, and highly influenced by the advisor’s expertise and experience. In contrast, robo-advisors have definite advantages.

First, robo-advisors provide a high level of customization. Through advanced algorithms and models, they can quickly and accurately analyze investors’ financial conditions, investment goals, and risk preferences, creating particular portfolios for each individual,no matter a short-term capital appreciation or a long-term retirement planning.

Second, robo-advisors are cost-efficient. Traditional advisory services often involve relatively high consultation and management fees, which can be a significant burden for individual investors. By an automated service process, robo-advisors apparently reduce operating costs. Generally, their management fees are about 0.5%–1% lower than those charged by traditional advisors, resulting in considerable savings over the long term.

Furthermore, robo-advisors offer real-time responses and convenience. Investors can access their portfolios anytime through mobile apps or website platforms and make transactions with ease. At the same time, robo-advisors continuously monitor market conditions and adjust portfolios in a timely manner to ensure that they remain aligned with the investor’s objectives and risk tolerance.

However, robo-advisors have some challenges,including data security and privacy protection, as well as concerns about the effectiveness and transparency of algorithms. Investors should carefully choose compliant and reliable platforms.

BDO Technologies Inc.

02/07/2026

01/30/2026

01/17/2026

01/10/2026

01/01/2026

12/29/2025

12/20/2025

12/13/2025

12/06/2025

11/29/2025

11/23/2025

11/22/2025

Address

Telephone

Website

Alerts

Shortcuts

Share

Category