Detailed narrative
Leadership Transition and Strengthened Management
Jack Abuhoff will transition to Executive Chairman, and Rahul Singhal will become President and CEO of Innodata, effective September 30. This planned transition aims to leverage Jack's focus on building long-term differentiating capabilities in enterprise and federal markets. Jayant Chauhan joined as Chief Financial Officer in July, strengthening the financial leadership team and focusing on capital allocation, capital markets, M&A, and investor communications.
AI Innovation as a Growth Engine
Research and innovation are increasingly pivotal, driving differentiation and expanding partnerships across the full model training life cycle, from pretraining and posttraining to model evaluation and benchmarking. This innovation is producing intellectual property and differentiation that generates demand, positioning Innodata as a strategic partner to leading AI builders.
Agentic AI and Enterprise Opportunity
Innodata has established an early position in agentic reinforcement learning, winning significant new programs covering personalization of long-horizon agents and reinforcement learning environments for desktop computer-use agentic tasks. The company aims to be the 'AI deployment assurance layer' for enterprises struggling to deploy AI agents with confidence, opening substantial opportunities in insurance and banking, with delivery already deepened with one big tech customer and started with another.
Benchmarking and Advanced Data Generation
The company released two public benchmarks designed to test frontier models on multi-turn, long-context, and multi-model interactions, surfacing failure modes that standard leaderboards miss. These benchmarking engagements lead to data strategy recommendations and scaled data generation to improve models, extending reasoning capabilities across five frontier labs and five domains.
Physical AI and Multimodal Data Collection
Innodata is building the required data and measurement layer for embodied intelligence, signing two research agreements with a leading university and committing to a motion capture lab expected to come online in the next few months⏳. This lab will collect sub-millimeter precision data for training robots and physical AI foundation models. Successful egocentric data collection pilots with robotics companies are leading to scoping enterprise-scale multimodal programs, including a multilingual speech program and a large egocentric program.
AI Cyber Training Suite for Secure Code
The first stage of Innodata's AI Cyber Training Suite was released, comprising 12 datasets and evaluation systems designed to train AI coding agents to write secure code and repair software vulnerabilities. Testing showed that fine-tuning open-weight models on a portion of this data more than doubled their repair rate for verified flaws, addressing a critical blocker for enterprise AI adoption related to trust in AI-generated code.
Structural Economics and Market Positioning
Innodata's business model demonstrates strong operating leverage, with adjusted EBITDA growing 1.6x faster than revenue. Margin quality is driven by a mix shift towards higher-value pretraining programs and off-the-shelf datasets where the company retains IP. Management believes Innodata is well-positioned to benefit from broader market dynamics, including debates around AI CapEx, commoditization of inference, and security incidents, as these trends all increase the need for data evaluation and assurance services.