Bhubaneswar: Indian entrepreneur Sam Pasupalak has introduced SKY-001, an artificial intelligence (AI) model designed to handle complex engineering design tasks, with its developers claiming that it can deliver comparable or better results than leading AI systems at a substantially lower cost.
Developed by the team at Skyfall AI, the model focuses on computer-aided design (CAD) and building information modelling (BIM). According to the company’s reported benchmarks, SKY-001 achieved a slightly higher score than OpenAI’s GPT-6 Astra on a specific CAD evaluation while costing around 150 times less per sample.
The model represents an effort to make AI more practical for specialised engineering applications, where accuracy, consistency and the ability to manage intricate designs are important. However, the performance claims are based on the company’s own tests and have not been independently verified.
SKY-001 aims to address complex engineering challenges
AI tools are increasingly being explored for tasks involving product development, design automation and technical problem-solving. While general-purpose models can generate code, descriptions and design concepts, translating those outputs into complete engineering designs can be challenging.
Complex products often consist of multiple assemblies, individual components and interconnected parts. Each component must meet specific requirements and fit into the overall design. A mistake in one section can affect the functionality of the entire product.
SKY-001 has been developed to address this challenge by breaking complex engineering assignments into smaller, manageable stages rather than attempting to generate the entire design in one step.
According to Skyfall AI, the model first identifies the major assemblies and individual components required for a project. It then develops a structured plan for constructing them, using visual representations, sketches and sequential construction steps.
This approach is intended to help the system manage projects involving hundreds or even thousands of individual steps while maintaining consistency across the design.
The company’s focus on specialised engineering tasks distinguishes SKY-001 from general-purpose AI assistants, which are designed to support a much wider range of activities.
How the model uses CAD and BIM
Computer-aided design is widely used to create and modify detailed digital representations of products, machinery, components and structures. Building information modelling, meanwhile, supports the planning and representation of buildings through digital models containing information about their physical components and layouts.
SKY-001 is designed to work across these two areas.
For CAD assignments, the model aims to organise a product into assemblies and individual parts before developing the necessary construction sequence. This staged approach is intended to reduce inconsistencies and ensure that the different components contribute to a coherent final design.
For BIM-related work, the model can take a building outline and divide it into individual units and rooms. Such a process requires the system to understand how separate elements fit within the larger structure.
Skyfall AI says its method allows SKY-001 to handle complicated projects without losing sight of the overall design. If demonstrated reliably in practical environments, this capability could be relevant to engineers, product designers, architects and other professionals working with detailed digital models.
Nevertheless, actual usefulness will depend on factors such as the complexity of the project, the quality of the generated design, compatibility with existing software and the need for human review.
SKY-001 versus GPT-6 Astra: what the reported figures show
Pasupalak shared benchmark results to explain how SKY-001 performed against other advanced AI models tested by his team.
One of the reported comparisons concerns the Intersection over Union (IoU) metric, which measures the degree of overlap between a generated result and a reference design.
SKY-001 reportedly achieved an IoU score of 79.7 per cent, compared with 79.1 per cent for GPT-6 Astra. The difference is 0.6 percentage points, indicating a relatively narrow gap in the specific evaluation.
The company also reported a significant difference in the cost of processing a CAD sample.
According to the figures cited in the report, SKY-001 costs approximately $0.005 per sample, while the corresponding GPT-6 Astra comparison is listed at $0.75. On those figures, running the task with GPT-6 Astra costs 150 times as much.
Pasupalak has also claimed that SKY-001 is between 51 and 125 times faster than the frontier models included in the team’s tests, with cost savings varying according to the model being compared.
These figures could attract attention from organisations that use AI for repeated or computationally demanding engineering work. Lower processing costs can be particularly valuable when projects require numerous iterations or involve large numbers of components.
However, benchmark scores and operating costs alone do not establish overall superiority. Results can vary depending on the task, evaluation method, hardware, model configuration and quality requirements.
The reported comparison should therefore be understood as a result from a specific test rather than proof that SKY-001 outperforms general-purpose AI systems across all applications.
Who is Sam Pasupalak?
Sam Pasupalak is an entrepreneur originally from Bhubaneswar, Odisha, who is currently based in San Francisco, the United States. He is the co-founder and chief executive officer of Skyfall AI, the company developing SKY-001.
Pasupalak completed his bachelor’s degree in mathematics and computer science at the University of Waterloo in Canada in 2011. He later pursued a master’s degree in political science and government at the London School of Economics, completing it in 2019.
His academic background at Waterloo included areas such as artificial intelligence, robotics and human-computer interaction.
Before establishing Skyfall AI in 2025, Pasupalak co-founded Maluuba in 2011 and served as its chief executive officer. The AI research startup was acquired by Microsoft in 2017.
His earlier experience in AI research and language understanding forms part of his professional background in the technology sector. The development of SKY-001 reflects a move towards specialised AI applications focused on technical and engineering workflows.
What SKY-001 could mean for engineering professionals
Specialised AI models could play a growing role in engineering by assisting with repetitive design work, organising complex projects and reducing the time needed to develop digital prototypes.
If SKY-001 can reproduce its reported results in independent evaluations, its combination of task-specific performance and lower operating costs could make it worth considering for certain engineering workflows.
Potential applications include product component planning, digital building layouts and the generation of structured design sequences. Such tools could help professionals explore alternatives more quickly, although they would not remove the need for engineering expertise, quality checks or compliance with relevant safety standards.
For businesses considering adoption, the key questions will include how the model performs on real projects, whether its outputs meet required tolerances, how easily it integrates with existing design tools and what the total deployment costs involve.
Independent testing will also be important in determining whether the reported speed and accuracy advantages remain consistent across different types of engineering problems.
Independent verification remains important
SKY-001 highlights the growing interest in AI systems built for specialised professional tasks rather than broad, general-purpose use. Its reported CAD results and lower per-sample cost offer an early indication of the approach being pursued by Skyfall AI.
However, the company’s benchmark figures have not been independently verified, and the reported results do not establish that the model is better than competing systems for every engineering application.
Further testing, practical deployments and transparent comparisons will help determine its reliability and commercial value.
For now, SKY-001 is a development to watch in specialised engineering AI. Its long-term significance will depend not only on benchmark performance but also on whether it can deliver accurate, dependable and cost-effective results in real-world design environments.
