QuanTech
Analytics in the era of cloud computing
QuanTech is an emerging area that combines quant (or analytics) with technology - offering an integrated approach to modelling, modern data engineering practices and cloud computing. Its aim is to first disrupt, second modernize and then streamline the delivery of quant/analytics services in a firm – saving time and cost and avoiding unnecessary hassle.
QuanTech is a cousin of FinTech and RegTech
- FinTech is an emerging industry that applies technology to improve activities in finance
- RegTech, is a new technology that uses information technology to enhance regulatory processes. Often regarded as a subcategory under FinTech, RegTech puts a particular emphasis on regulatory monitoring, reporting and compliance and is thus benefiting the finance industry
- Similarly, QuanTech is an emerging framework and set of practices that uses technology to improve analytics (in the heavily regulated financial sector and also wider). All models are part of the analytics including AI and ML models
A few practical benefits and examples follow:
01
QuanTech and QuantOps make the modelling process iterative, fast and easy to operationalize in production.
02
QuanTech encourages the use of practices like CI/CD which provide automation in model development, validation and implementation. Automated testing is an integral part of Quantech that helps in continuously assessing and assuring the validity of models with known datasets.
03
Use of DevOps practices creates a common infrastructure so that model development and validation take place in an environment that is as close to production as possible.
04
Data quality checks are automated to continuously verify data lineage and pinpoint transformation errors. This helps in significantly reducing downstream problems that require time consuming investigations.
05
Quantech encourages the practice of wrapping quant models with microservices. Models can be built using the best languages and frameworks that suit the need and still be integrated with other parts of a large application using this approach (e.g. new credit risk models may be in Python, dated trading book models in C++ and both libraries need to be “called” in a stress testing application).
06
Containerization (e.g. Docker) and infrastructure orchestration (e.g. Kubernetes) provide a standardized, scalable environment for deployments while also reducing errors due to package management, versioning and software dependency management.
QuanTech in detail
QuanTech is an emerging area that combines quant (or analytics) with technology - offering an integrated approach to modelling, modern data engineering practices and cloud computing.
1
Modelling/ ModelOps/ MLOps
- Modelling comprising the full lifecycle from build, validate, implement, productionise, monitor to risk manage and govern. It includes all models (traditional, AI, ML)
- ModelOps is the operationalization of all models, i.e. efficient lifecycle implementation
- MLOps is the operationalization of ML models and hence is a subset of ModelOps. In some contexts MLOps is used to include the entire lifecycle of ML models
2
DataOps/Data engineering
- DataOps is used to describe a process-oriented methodology and its automated implementation with the aim of improving the entire data lifecycle from data preparation to reporting
- One goal is to shorten the cycle time of data analytics in alignment with business goals
- DatOps recognizes the dependencies between the two teams of data analytics and technology operations
3
Software architecture
- Software architecture describes a set of aspects and decisions that are important to a software, i.e the organization of the system, how the system parts communicate with each other, external dependencies, guidelines and implementation technologies, risks, performance, security, etc.
- It is a blueprint of the software and defines how the software will function. It becomes a basis for communication between the stakeholders
4
DevOps
- DevOps is a set of practices that combines software development (Dev) and IT operations (Ops)
- DevOps focuses on continuous delivery by leveraging on-demand IT resources and by automating test and deployment of software. This merging of software development and IT operations improves velocity, quality, predictability and scale of software engineering and deployment
- Several DevOps aspects come from the Agile methodology
1. Example of Modelling/ModelOps
All modelling activities can be categorized into seven phases.
The figure below shows the full model lifecycle with model governance at its centre.
This lifecycle applies in theory to all models. However, the different phases are applied to varying degrees depending on the regulatory requirements. For example, the full lifecycle is applied most rigorously to capital models.
Tap to
zoom image
zoom image
2. Example of DataOps/Data engineering
This diagram shows a simplified view of an Model Risk Quantification (MRQ) implementation. The calculation component implements the methodology but for a usable solution needs to be linked to a wider infrastructure.
Tap to
zoom image
zoom image
3. Example of software architecture
(stress testing tool)
Tap to
zoom image
zoom image
4. Example of DevOps
Tap to
zoom image
zoom image
A classic example of QuanTech: enterprise stress testing solution (Pluto)
There are many real world situations that require all four aspects of QuanTech to work in unison. A classic example is the delivery of firm-wide stress testing orchestration tool in a large financial institution requires:
- Thousands of models and an efficient model lifecycle
- Data ETL from a large number of disparate sources and with frequent updates
- The different sub-systems (trading book, mortgages, credit cards, central risk, …) need to communicate with each other and manage dependencies
- Continuous integration and delivery (CI/CD) and automation are required to able to provide the required speed to users.
Get in touch
© Basinghall Analytics 2022. All rights reserved.