Introduction
Solvendo has been working with machine learning since 2018. Through this journey, we have made multiple philosophical deep dives to better absorb the continuously evolving nature of AI/ML and the profound effect it continues to have on our world.
In Rem Musings: Wisdom from the Vedas
According to Hindu or Vedic belief, Shrutis are verses passed down through generations through recitation. The Vedas, according to belief, are revelations that the great seers experienced as Darshana (divine vision), when they were in a state of total passive openness to nature.
To our understanding, all interpretations are Paurusheya, or human-centric, while the Vedas are of non-human origin. In the interpretation of the Vedas — even that of the seer — between the words there will remain a gap, as interpretations shall always be conditioned by the cultural and intellectual context of the interpreter.
Humans express themselves through words and are thus limited by them; words are often imprecise. This permits multiple connotations and interpretations depending on context.
Most religious texts attribute divine origin to their text; this approach makes adoption simpler and less debatable.
Common law, another centuries-old data set that continues to evolve through constant human input and iteration — namely statutes and judge-made law — is perhaps another good example.
Common law permits constant interpretation of the law by human oracles. The oracles are judges in pursuit of the cause of justice. The task of interpreting law entails interpretation of words and language in the context of facts, legislative intent, and justice principles. It is in fact the interpretation of words written by one set of humans, by another set of humans. This process, being inexact, has been a source of innovation and also the cause of significant inefficiency.
Machine Learning: Bridging Divine Wisdom and Human Iteration
Between the divine origin model and models like common law that require constant human regurgitation lies Reinforcement Learning and generative pre-trained transformer-based models.
The speed, efficiency, accuracy, and intelligence of foundational models are no longer debatable. They have arguably, in some ways, reached the point of ‘singularity’.
Machine learning models (Reinforcement models) are likely to assist human effort in architecting better systems and in intelligently automating these processes through continuous reinforcement learning, ensuring they are more scalable compared to human interpretation. Human oversight will ensure the continuous learning loop remains in line with the human intent and justice principles.
Solvendo.io: Architecting the Future of Credit Markets
Solvendo spent five years architecting and building technology infrastructure that will drive the future of credit markets globally. We continue to build our proprietary reinforcement model that works seamlessly with this infrastructure, purpose-built to work with financial institutions.
Combined, the solution will provide real-time data and intelligence to stakeholders, making capital more accessible and affordable. We achieve this by reducing information asymmetry.
Most importantly, the solution is an insurance policy for small businesses, which we believe can be the next growth engine for the global economy.
Regional Implementation and Global Ambitions
We have begun implementing our solution in India, currently the third largest economy — one that promises unlimited potential given access to affordable credit.
We hope to be able to launch our solution in the United States and other jurisdictions by the end of 2026.
We believe our solution will demonstrate what AI and foundational models can achieve when applied to real financial problems — transforming credit markets and expanding access to capital globally.