Amazon CodeWhisperer, the AWS AI coding companion, is a step change in developer productivity tools. Based on generative AI technology, Amazon CodeWhisperer offers contextualized code snippets or recommendations based on natural language prompts to build software quickly, responsibly, and securely. It enables productivity gains and increases accuracy for accelerated digital transformations. Amazon CodeWhisperer ensures enterprises have greater control over AI-generated code, especially the code written by developers who may have a limited understanding of code attribution, quality, and security requirements.
Persistent Systems, a global digital engineering provider, has run several pilots and formal studies with Amazon CodeWhisperer that point to shifts in software engineering, generative AI-led modernization, responsible innovation, and more. This post highlights four themes emerging from Persistentâs Amazon CodeWhisperer experiments that could change software engineering as we know it.
Beyond productivity gains: Reimagining coding with Amazon CodeWhisperer
In this section, we discuss some of the ways that Amazon CodeWhisperer is reimagining coding.
Improving responsible delivery
Ownership, explainability, and transparency of AI-generated code are the most contentious points for the commercial adoption of coding companions such as Amazon CodeWhisperer. Amazon gives developers complete ownership of the code they write using Amazon CodeWhisperer. The Amazon CodeWhisperer team has carefully curated the training data and omitted restrictive licenses, ensuring developers donât inadvertently use restrictively licensed code when they use Amazon CodeWhisperer. In addition, because recommender pipelines can be strongly influenced by open-source code, if Amazon CodeWhisperer detects a lineage, it flags the license references (for example, MIT or Apache, an open-source project). This enables the developer to attribute code snippets to the source owners, instituting coding best practices. Although Amazon collects data such as code snippets, recommendations, and comments from files open in the integrated development environment, for Amazon CodeWhisperer Professional users, these are not stored or used to train the model. Also, Amazon CodeWhisperer Individual users can opt out of sharing content with AWS, limiting the chances of this being reproduced as recommendations to other users.
Persistentâs approach to generative AI mirrors Richard P. Feynmanâs thinking, who said, âI would rather have questions that canât be answered than answers that canât be questioned.â Persistent prioritizes responsibility, accountability, and transparency to build client trust. One example of the potential of Amazon CodeWhisperer lies in its ability to reference code, helping clients circumvent legal liabilities that could derail other rewards. For more information about Persistentâs approach to generative AI, refer to Generative AI Services and Solutions.
Moving code security upstream and upfront
Seasoned developers will tell you that security cannot be tested-in; it must be built from the ground up. Although some approaches, such as DevSecOps, make it easier for developers, code security experts, and operations teams to embed security testing while the code is written, Amazon CodeWhisperer takes this one step further. It runs security scans on the code directly in the integrated development environment (IDE), allowing a single developer resource to test the code for quality and security. This highly automated, shift-left scenario for security testing enables enterprises to arrest defects upstream and remedy them at a fraction of the cost and time. Especially now, when coding, with the advent of generative AI moving closer to business users, the automated, in-line security scans in Amazon CodeWhisperer will provide less rework, faster time to production, and resilient code.
Persistent helps leading global organizations fortify their business applications with code embedded with security guardrails. It believes security testing has to shift closer to the developer (professional or citizen) and be encoded into applications as they are written. Amazon CodeWhisperer, with its transformative power to fast-track not just coding but secure coding, fits well into the narrative.
Enabling developer skills to undergo a reboot
Most developers must undergo at least 4 months of training before being tagged to projects. In our pilot, Amazon CodeWhisperer condensed the training period to 1 month with reduced cognitive load concerning understanding the context or coding language. We see this bearing on how companies hire developers, evaluating not the coding knowledge, which has been largely abstracted, but on the prompt engineering expertise and the ability to be creative with tools such as Amazon CodeWhisperer.
The parameters for professional developers will change, and quickly depending on their ability to tune the input to get the desired answer. This also opens the field for citizen developers or business technologists, bringing coding closer to the business.
Driving implementation closer to strategy
With so many moving parts, businesses and their technology partners will return to the whiteboard together. The engagement model will evolve to factor in these new variables (such as faster coding timelines, secure code, more citizen developers, or domain-oriented developers) unleashed by Amazon CodeWhisperer. Coding will now move closer to the business, automatically incorporating security guardrails and mandatory regulations into software applications as they are written, all at scale. And with verticalized workloads, success will depend on the development teamâs domain expertise and the ability to translate code into innovation. This means the implementation of the companyâs vision through this code will become even more watertight because it adheres to strategic pillars of security, quality, and speed.
From long shots to offshoots â what the future holds
We extrapolated these themes to map a future where Amazon CodeWhisperer can help realize âdelivery moon shotsâ that, up until now, were aspirational. The future looks something like this:
- Zero-wastage â Amazon CodeWhisperer, especially with its proactive security scans and reference tracker tool, will ensure the code is of shippable quality, enabling every allied functionâfrom business to developersâto add value and minimize wastage in terms of effort, time to value, or rework. This will bring a singular focus on the core job for each stakeholder, further enforcing a value-first mindset.
- Zero ramp-up â The ability to support multiple coding languages, factor in developer notes and comments into code suggestions, and offer lines of code on the fly makes Amazon CodeWhisperer the perfect antidote to the cold start problem for developers. As mentioned, developers donât need a gestation period before being onboarded on a project. This dramatically cuts down the time to value, allowing implementation partners to deploy resources across projects for better monetization dynamically.
- Zero-shot translation â Amazon CodeWhisperer supports multiple programming languages, such as Python, Java, JavaScript, TypeScript, SQL, and more. It will be able to translate code from one programming language to another, or what is called zero-shot translation ability, where it uses reference code in language A to write code in language B more accurately. This unleashes significant changes in how legacy modernization projects are planned and implemented. With the zero-shot translation ability of Amazon CodeWhisperer, Persistent is confident legacy modernization will become faster and no longer be a moon shot.
- Zero lifting â Amazon CodeWhisperer is optimized to generate accurate code for other AWS offerings, such as Amazon Simple Storage Service (Amazon S3) and Amazon DynamoDB. The accurate code generation makes the lift easy. Because AWS and other major cloud service providers are now pushing forward a multi-cloud narrative, Persistent expects Amazon CodeWhisperer to improve accuracy while recommending code for other solutions offered by AWS peers. This makes the road smoother for multi-cloud or multi-platform settings, eliminating the heavy lifting required while shifting workloads from one service vendor to anotherâsupercharging digital transformation 2.0.
Conclusion
Amazon CodeWhisperer goes beyond improving developer productivity: it democratizes coding and brings it closer to business users while ensuring best practices such as code attribution and enhanced security are never out of the purview.
Persistent is excited about Amazon CodeWhisperer and its potential impact on businesses and partners. It is working to create an Amazon CodeWhisperer-ready developer workforce and alerting its customers about its benefits to drive adoption. Persistentâs strong partnership with AWS makes it the best-fit technology partner to help businesses capitalize on the intrinsic value of Amazon CodeWhisperer.
To learn more about Persistentâs generative AI philosophy that reimagines the way software is engineered today and how Amazon CodeWhisperer aligns with it, refer to Generative AI Services and Solutions.
About the authors
Dr. Pandurang Kamat is Chief Technology Officer, responsible for advanced technology research focused on unlocking business value through innovation at scale. He is a seasoned technology leader who helps customers improve user experience, optimize business processes, and create new digital products. His vision for Persistent is to be an innovation powerhouse that anchors a global and diverse innovation ecosystem, comprising of academia and start-ups. He holds a bachelorâs degree in Computer Engineering from Goa University and Ph.D. in Computer Science from Rutgers University. He is a well-published author with several international research publications, an ACM-India Eminent Speaker, serves on the board of studies at universities, and mentors technology start-ups.
Ankur Desai is a Principal Product Manager within the AWS AI Services team.
Kiran Randhi works for Amazon Web Services as a Principal Partner Solutions Architect in Seattle, Washington. He works closely with AWS Global Strategic SI partners to develop and implement effective cloud strategies that allow them to fully leverage the benefits of cloud technology. Kiran helps CIOs, CTOs, and architects turn their cloud visions into reality by providing architectural guidance and expertise throughout the implementation of strategic cloud solutions. He focuses on AWS security, Migration & Modernization, Data & Analytics, and other technologies to build solutions for different industries in the cloud.
