An AI engineering company
We build products and ship systems for clients who need both. Currently focused on quantitative finance and machine learning, with data engineering and software engineering services alongside.
Featured engagement · 2024 — present
A research and execution platform for algorithmic trading strategies in frontier and emerging markets. Includes data ingestion, feature engineering, backtesting, walk-forward validation, and live execution components.
Part of what we do
Quant · ML
Algorithmic trading research, regime detection, and execution infrastructure for frontier markets.
AI Engineering
Production ML systems, agent platforms, evaluation pipelines, and RAG systems at scale.
Data
Pipelines, warehouses, and real-time analytics for organizations that need their data working harder.
Research
Selected open work on systems we build, published as we ship them.
Selected work
A research and execution platform for algorithmic trading strategies in frontier and emerging markets. Includes data ingestion, feature engineering, backtesting, walk-forward validation, and live execution components.
Built and deployed a production retrieval-augmented generation system serving 10M+ daily requests. Architected the embedding pipeline, vector search layer, and response generation system. 40% latency reduction and 99% query resolution rate.
Lead AI Engineer for a HIPAA-compliant therapy platform built on LangGraph. Designed multi-agent therapy session flows with safety guardrails, clinical-evaluation pipelines, and audit trails meeting healthcare-compliance requirements.
What we believe
We build code and relationships meant to last. Restraint over flourish, working systems over impressive demos, honest reporting over optimistic estimates.
Latest writing
A founding note. Why we exist, what we're trying to build, and what we mean by 'restraint over flourish.' Written for our future selves to keep us honest.
Read post →Lessons from architecting a RAG system serving 10M+ daily requests. How to reason about end-to-end latency, what to optimize first, and where the surprising bottlenecks usually are.
Read paper →Cataloging reward-hacking strategies observed across 200+ trajectory annotations on Terminal-Bench. We propose a taxonomy and recommend evaluation patterns that resist gaming.
Read paper →Get in touch