Practice area · Kansas City and remote
Agents can now run most of the machine learning workflow: pulling and cleaning data, writing feature code, fitting candidate models, running backtests and watching for drift. Speed is the easy part. The hard part is knowing when an agent's model is wrong, because an agent that is 95% reliable per step completes a 20-step pipeline only about 36% of the time. Vector 1 designs agentic ML workflows where every stage is verified against a known answer before the next one starts.
Tell us what you're trying to measure, forecast or decide. We work with teams in Kansas City and remotely, and we reply within two business days.
Email [email protected]Why a 95%-accurate agent fails most 20-step tasks, and what a verifier, retries and human escalation do to success rates and cost per completed task.
Open demo →Discover a causal graph from data and see how far the naive correlation is from the real effect of speeding up a page.
Open demo →Self-hosting an LLM costs $10.42 or $1.49 per million tokens on the same hardware. The difference is utilization; find your break-even against API pricing.
Open demo →
A technical walkthrough of how machine learning, causal inference, and autonomous agents combine into a production marketing intelligence stack — with code,…
By Brian Curry | Vector1 Research “The question is never ‘which model is best.’ It is ‘which model is best for this series, at this horizon, under this regime,…
By Brian Curry | Vector1 Research “An agent without memory is not an agent. It is a very expensive stateless function.” The dominant approach to AI agent…
Brian James Curry · Vector1 Research Fourth in a series on the layer around the model. “You can’t test quality into a product.” — a maxim from manufacturing…
An agent that runs for more than a few minutes has a memory problem. Every tool call, every observation, every intermediate result is something it might need…
In the previous article, retrieval was a one-shot transaction. A user asked a question, the system pulled context, the model wrote an answer, the conversation…
The machine learning lifecycle has traditionally been a labor-intensive process requiring significant human intervention at every stage — from data preparation…
How autonomous AI research agents are discovering novel algorithms that surpass human-designed solutions This article is inspired by and draws concepts from…
The way we think about machine learning at scale has fundamentally changed. We’ve moved past the era of singular “hero” models doing everything alone. Today’s…
Machine learning work carried out by AI agents: preparing data, writing feature code, fitting and comparing models, backtesting and monitoring, directed and reviewed by a person. The agent does the execution; the verification and the judgment are designed in.
Only with checks at each step. Errors compound across a long pipeline, so the reliable approach is to test every stage on data where the right answer is known, and to stop the pipeline when a check fails rather than passing a bad model forward.
Yes. Vector 1 is based in Kansas City and works with teams in the metro and remotely.