Instead of hiring actors and a film crew for every ad, Oxus lets you type what you want and an AI creates a realistic video ad with a virtual spokesperson in minutes.

Oxus's video ad platform leverages state-of-the-art text-to-video and text-to-speech generative models to produce broadcast-quality advertisements featuring photorealistic AI-generated actors and voices. Users define campaign parameters through a node-based automation interface, which orchestrates the generation of dozens of creative variations optimized for different audiences, platforms, and messaging angles. The system eliminates traditional production bottlenecks—casting, filming, editing—replacing them with prompt-driven workflows that output publish-ready creatives in minutes rather than weeks. With over 1,000 ads published for 100+ businesses, the platform has demonstrated measurable ROI within days of deployment, and its hands-on AI specialist support ensures that non-technical marketing teams can fully leverage the technology. The architecture supports continuous iteration, allowing performance marketers to rapidly kill underperforming variants and scale winners without returning to production.

It's like having a Hollywood studio in your laptop, except the actors never complain, never need coffee breaks, and can speak 20 languages before lunch.

Instead of auditors spending weeks reading thousands of documents to check if a company follows the rules, Oxus's AI reads everything in minutes and flags exactly what needs attention.

Oxus's YC W26 audit automation platform applies large language models and advanced NLP to the traditionally manual, labor-intensive process of internal SOX compliance auditing. The system ingests complex corporate documentation—policies, process narratives, control descriptions, and test evidence—and uses fine-tuned document understanding models to automatically scope audits, map controls to risks, generate testing procedures, and produce review-ready workpapers. What previously required weeks of senior auditor time and expensive outsourcing to Big Four firms is compressed into minutes of AI-driven analysis. The platform offers enterprise-grade security with customer-hosted deployment options, addressing the data sensitivity concerns inherent in audit work. Founded by three MIT alumni, the team brings deep technical expertise to a domain historically underserved by technology, targeting pre-IPO and public companies where SOX compliance is both mandatory and costly. The direct founder engagement model ensures rapid iteration based on practitioner feedback.

It's like replacing an army of accountants with magnifying glasses with one super-accountant who has X-ray vision and never needs to sleep.

Oxus's AI watches which ads perform best, figures out why, and automatically makes the next batch of ads smarter—like a marketing team that learns from every single click in real time.

Beyond generating video ads, Oxus closes the creative feedback loop by applying machine learning to the performance data generated across its 1,000+ published ads and 100+ client campaigns. The system ingests engagement metrics—click-through rates, view completion rates, conversion events, and cost-per-acquisition data—from integrated ad platforms and correlates them with creative attributes: actor appearance, voice tone, script structure, visual composition, call-to-action placement, and audience segment. Pattern recognition models identify which creative elements drive performance for specific verticals, audiences, and platforms, and these insights are fed back into the generative pipeline to bias future ad creation toward high-performing configurations. This creates a compounding intelligence advantage: each new campaign makes the system smarter, and clients benefit from cross-portfolio learnings without exposing proprietary data. The node-based automation architecture allows these optimization signals to be embedded directly into production workflows, enabling what is effectively autonomous creative iteration guided by real-world performance data.

It's like having a focus group of millions that votes with their thumbs, and an AI director who rewrites the script between every take based on the applause.