I architected and led the implementation of an agentic system that automates the dealflow process for 1kx. It orchestrates several layers of agents, starting from deep research and data augmentation to specialized agents that each score the deal along a specific dimension up to a final holistic agent that takes all other agents’ outputs and distills them to a final recommendation. A critical aspect of the system is the measurement and evaluation of quality to enable data-driven decision making and observability into the system down to span-level. For this I have built the 1kx AI Platform and AI Lab, which are foundational capabilities that unlock managing, optimizing and iterating on AI systems at scale. It includes capabilities around LLM observability, cost tracking, prompt management, prompt versioning, eval dataset management, experiment orchestration, automated evaluations, backtesting and A/B testing of system configurations consisting of LLM model, parameters and prompts.
The implementation of the foundational AI Platform capabilities enable the fund to build, measure and optimize AI at scale in a data driven way. With AI Dealflow I have also implemented the first high-impact use case leveraging the foundational capabilities to build an intelligent and self-improving system.