Aditi Purandare is an Analyst at Hitachi Ventures, a $1B AUM independent financial fund of Hitachi. She focuses on investing across the enterprise software stack, from the infrastructure layer to the application layer. Prior to starting at Hitachi Ventures, Aditi conducted research in applied Machine Learning and data science across a range of domains, including a Fulbright exchange in AI-enabled portfolio optimization. She holds a Computer Engineering degree from Northeastern University.

What is a value, principle, or philosophy that has meaningfully shaped your journey? How does it continue to influence how you invest, lead, or approach challenges?

Intellectual curiosity has been a key driver in my journey. As a student interested in research (I would have been in an applied AI PhD program if I wasn’t an investor), I found a great deal of excitement in conducting literature reviews and filtering those approaches for feasibility and relevance. Presently, in venture, this looks more like seeking out new promising technologies and markets, while identifying challenges and opportunities in those spaces.  In a world where technology is evolving rapidly, intellectual curiosity also grounds me. On one hand, it encourages me to look broadly, identifying areas that I may have overlooked otherwise. On the other, it pushes me to dive deeper and iteratively filter out for the most relevant information.

As an investor, one of the most rewarding aspects of my job is being on the front lines for innovation and witnessing new breakthroughs. To best support founders and operators I speak with, I strive to channel my curiosity to be open-minded and approach each opportunity with genuine excitement about what is possible.

Is there a technology, idea, or movement that most people overlook (or dismiss) which you believe will define the next era of innovation? What draws you to this perspective?

While AI agents are undoubtedly ubiquitous, in my experience, well-architected agents are much less common. I believe that the future of enterprise AI will be driven by the latter. Among the many factors that contribute to a high-quality agent, one of the features that I view as particularly critical is the reasoning engine. Reasoning engines are the brains of agentic systems, carrying out the decision-making. However, using out-of-the-box LLMs, or even fine-tuned variants, does not effectively achieve this sufficiently.

A truly effective reasoning engine that understands user intent, decomposes complex tasks, and creates execution plans for each subtask, all while aligning reasoning to organization and persona, is important. Achieving this requires deep integrations into company data, continually updated semantic graphs, multi-step reasoning capabilities, RLHF, and self-healing mechanisms. These approaches collectively enable agents to operate with precision and adaptability.

Of course, well-architected reasoning engines alone are not the only crucial component for a well-functioning agent. High-quality agents need to have proper management for long-term and short-term memories, security built in for all potential agentic attack surfaces (including other agents, tools, memory, reinforcement learning, prompts), orchestration, observability/explainability, and more. The reasoning engine, however, is a natural starting point as it drives the agent’s cognition. The next era of agentic innovation will rely on systems that integrate reasoning, memory, security, and adaptability for reliable and accurate solutions.