Sanat Anand

Source: Notion | Last edited: 2023-10-30 | ID: 0551f52a-b05...

---

File attachment

Video

What’s needed from Sanat Anand:

1. An outline of the ideal situation including the prospective timeline for the preparation before launching the system LIVE.
2. Details of the resources that might be needed for the completion of the tasks.
3. Justifications for the time period quoted for the completion of the preparation.
4. Details about the kind of partnership the candidate is seeking in terms of base pay, bonus pay, and other benefits.
5. Thoughts on the necessity of starting from scratch, and concerns over reusing or repurposing existing IP.

2023-07-16 Meeting Memo with Sanat Anand

Background and Experience

• Sanat Anand has an undergraduate degree in computer science from IIT Bombay and worked at WorldQuant and Citadel in equities trading.
• They then moved to Amber Group in Singapore to build crypto strategies based on NLP.
• They developed their own data using social media platforms and trained their own NLP models.

Performance and Results

• Sanat Anand’s NLP strategies have low correlation with existing CTA strategies and produce good returns.
• Their out-of-sample Sharpe ratio is between 3 and 3.5.
• Their strategies can handle a capacity of around 50 million dollars and generate returns of 30 to 35%.
• Adding NLP strategies to existing CTA strategies can increase the overall portfolio’s Sharpe ratio.

Trading Frequency and Turnover

• Sanat Anand’s strategies capture sentiment at a half-hour frequency and immediate events on Twitter at a one-minute frequency.
• The overall turnover of capital is around 40-45%, with a holding period of more than two days on average.
• The trading frequency is one minute for immediate events and half an hour for sentiment.
• The strategies work as a portfolio, with some trades happening every minute and others every half an hour.

Nature of Strategy and Data Collection

• Sanat Anand’s strategies are based on NLP models trained using machine learning, specifically fine-tuning pre-trained models like BERT and Beta on crypto data.
• They collect data from various social media platforms in a specific way, assigning weights to it.
• The IP of their strategies lies in the data collection method and their NLP model.

Scalability and Portfolio

• Sanat Anand’s strategies are scalable and can be run with a portfolio of around 40-50 tokens.
• The strategies work together in tandem as a portfolio.
• The turnover of capital is around 40-45% daily, but the entire portfolio is not traded.
• Trading frequency amounts to about two to three trades per day.

Impact of External Factors

• Amber Group, the firm the Sanat Anand works for, was affected by external factors like the FTX crash.
• The firm decided to downscale and reduce capital in quant strats, affecting the Sanat Anand’s bonuses.
• Sanat Anand feels that their strategies are independent and can scale well to another place.

Trading Strategy Setup

• The trading strategy requires a significant amount of setup, including building models and gathering data.
• The setup process typically takes several months.

Employment Relationship and Ownership

• The IP and source code are owned by the current employer (Amber Group).
• If a new employment relationship is formed, the work will need to be rebuilt from scratch.

Employment Structure

• There are two ways of working together: straightforward employment or as a third party researcher.
• In the employment structure, the base pay is evaluated based on an imaginary AUM.
• The bonus pay is 20-30 percent of the base pay.
• For third party researchers, a template is available.

Timeframe for Trading Strategy Implementation

• The setup process typically takes several months and cannot be rushed.
• It is necessary for successful implementation of the trading strategy.

Technical setup and infrastructure

• The project requires setting up data scraping and using the Twitter API for data collection.
• AWS servers are needed to store the data.
• A trading engine and backtesting engine are necessary.
• The capacity for the project is around 50 million.

Building models

• Training the models requires time and expertise.
• The researcher plays a crucial role in model development.

Past experience with Citadel and Amber Group

• Employment relationships were established with both companies.
• No immediate arrangement or capital constraint existed.
• Project performance determined yearly bonuses.

Protection and unique aspects

• No NDA or non-compete agreement is in place.
• Trading with NLP is not copyrightable, but code can be.
• The value comes from the details and nuances of the project.

Discussion and information sharing

• Further details and information will be shared in written form.
• Timeline, helper requirements, and infrastructure will be discussed.
• The employer can provide specific details for better understanding.

Incentives and capacity

• The employer is incentivizing strategies with high returns.
• Orthogonal strategies are desired.
• The employer already has high-capacity machine learning strategies.

Orthogonal Strategies for Complementing Eon’s Strategies

• Orthogonal strategies provide backup when current strategies are stagnant.
• Low correlation with other PMs’ strategies ensures complementarity.

Setup Phase for Profitability

• The setup phase is necessary to make the trading strategy profitable.
• Specific resources like AWS servers, backtesting systems, and GPUs are required.

Time and Contextual Information

• Deploying the system may require more than two weeks.
• Understanding the reason for the time required is essential.