From Art Curation to AI: An Unlikely Path to Accessibility Innovation
Some of the most consequential technology entrepreneurs arrive at their missions through unexpected corridors. Mousumi Kapoor, Founder and CEO of Continual Engine, embodies this pattern precisely. With electrical engineering degrees from IIT Delhi and the University of Minnesota, followed by technology leadership roles at GE and an active practice in art curation, Kapoor has assembled a vantage point that few in the accessibility space can claim. That interdisciplinary lens—merging rigorous technical training with aesthetic judgment and enterprise operational discipline—has proven essential to solving one of digital education's most stubborn problems.
The problem itself is deceptively simple to state yet maddeningly complex to solve. Legacy higher education publishers and digital media platforms have long treated accessibility as a compliance checkbox, an afterthought bolted onto finished products. This approach doesn't merely court legal risk; it systematically excludes millions of learners with disabilities from accessing educational content. The traditional remedy—manual alternative text generation—collapses under the sheer volume and complexity of modern educational materials. A single chemistry textbook might contain thousands of intricate diagrams, mathematical notations, and chemical equations that resist straightforward description.
The Engineering Mountain Behind Automated Alt-Text
What Continual Engine has built under Kapoor's leadership represents something far more ambitious than conventional image recognition. The platform has grown into a $5M+ ARR business precisely because it tackles content that most AI systems simply cannot parse. STEM diagrams, chemical equations, and mathematical notations demand contextual understanding, not merely pixel-pattern matching. A neural network must recognize that a benzene ring diagram carries different semantic weight depending on whether it appears in an organic chemistry chapter or a materials science context.
Enterprise trust and model precision ultimately outweigh pure speed when building accessibility infrastructure that institutions will actually adopt.
Kapoor breaks down the engineering breakthroughs required to bridge this gap, detailing how deep learning architectures must be trained to extract hierarchical meaning from visual scientific content. The operational challenges of moving from manual workflows to AI automation extend beyond technical architecture. Each false positive in alt-text generation doesn't merely degrade user experience—it potentially violates the very compliance standards the technology exists to satisfy. This explains why Kapoor emphasizes that enterprise trust and model precision ultimately outweigh pure speed when building accessibility infrastructure that institutions will actually adopt.
Designing for the Margins, Upgrading the Center
One of the most counterintuitive insights Kapoor brings to the conversation concerns the relationship between accessibility design and general user experience. The prevailing assumption treats accessibility features as concessions to regulatory pressure—necessary overhead that subtracts from mainstream product development. Kapoor inverts this logic entirely. Designing for accessibility, she demonstrates, functions as a universal upgrade for overall user experience. The same structural improvements that enable screen readers to navigate complex content also enhance discoverability, searchability, and cross-platform rendering for all users.
This principle carries strategic weight for EdTech founders weighing resource allocation decisions. Accessibility investment positioned as compliance burden becomes a candidate for minimal viable compliance. Reframed as foundational infrastructure, it attracts the engineering talent and architectural attention that produces durable competitive advantage. Continual Engine's embedded API strategy reflects this understanding: by integrating accessibility generation directly into content production pipelines, the company enables material that is "born accessible" by default rather than retrofitted as final-step remediation.
Content born accessible by default eliminates the retrofitting bottleneck that has trapped legacy publishers in cycles of expensive remediation.
Navigating the Startup-to-Enterprise Transition
Kapoor's own trajectory with Continual Engine illuminates a transition that destroys many promising ventures: the shift from unstructured startup agility to structured enterprise scale. Early-stage companies prize velocity and adaptability, often at the expense of documentation, reproducibility, and governance. Enterprise customers—particularly universities and major publishers managing vast content libraries—demand the opposite. They require audit trails, service level agreements, and integration pathways that preserve their existing workflows.
The operational playbook Kapoor has developed centers on embedding rather than displacing. Continual Engine's APIs don't demand that customers abandon familiar content management systems; they layer accessibility generation into established production chains. This architectural choice reduces adoption friction while progressively demonstrating value. It also builds the operational muscle memory that sustains growth beyond initial pilot contracts.
Key Takeaways for Founders
1. Interdisciplinary foundations create defensible technical moats. Kapoor's combined background in electrical engineering, corporate technology leadership, and art curation equipped her with the cross-domain fluency necessary to tackle problems that specialists in single disciplines consistently miss.
2. Model precision and enterprise trust must precede scaling velocity. In compliance-sensitive domains, a single inaccurate alt-text generation can undermine an entire customer relationship. Founders should optimize for reliability metrics that enterprise buyers can defend internally before pursuing growth at all costs.
3. Accessibility design compounds as universal user experience improvement. Rather than treating accessibility as segmented feature development, founders should recognize that structural accessibility enhancements typically elevate product performance for every user segment.
4. Embedded API architectures smooth the startup-to-enterprise transition. Building integration pathways that complement rather than replace existing customer systems reduces adoption friction and builds the operational infrastructure necessary for sustainable scale.
This conversation offers an indispensable operational playbook for EdTech founders, enterprise AI product leaders, and technology executives who recognize that compliance mandates, properly understood, become competitive infrastructure. The learners who benefit from accessible content number in the millions. The organizations that build that accessibility into their foundational architecture may prove equally numerous in seasons ahead.
