Financial reasoning
Evaluation of how advanced language and reasoning models interpret market data, economic releases, policy signals, company information, and noisy real-world evidence.
Frontier model systems for finance and macro research
Kernel Research
Kernel Research develops frontier AI systems for financial markets and macroeconomic analysis. We build multi-agent research workflows, evaluation harnesses, and model-driven tools for reasoning under uncertainty across market, policy, and economic data.
01 Frontier Models
02 Agent Workflows
03 Macro Analysis
04 Risk Evaluation
Research areas
Kernel research program
Kernel research program
Frontier models, agent workflows, and market systems examined through testable hypotheses.
Research workflows
Agent systems that divide research tasks across hypothesis generation, data retrieval, quantitative testing, critique, and report synthesis.
Risk and uncertainty
Protocols for measuring confidence, drift, exposure, counterfactual behavior, and failure modes before model outputs are used in research workflows.
Method
Research-driven by design
Research-driven by design
Research systems built around testable financial and macroeconomic questions.
Our work combines machine learning, quantitative research, macroeconomic analysis, market microstructure, and research software engineering. We study how frontier models form assumptions, use tools, coordinate with other agents, and adapt as information changes.
The goal is not a single benchmark score. It is a reproducible research process for evaluating AI systems against real financial questions, explicit assumptions, observable data, and documented failure modes.
The research program is informed by open work on financial LLMs, reproducible finance environments, and collaborative autonomous research systems.
00Research question and economic context
01Dataset construction and source reliability
02Model and agent workflow design
03Backtesting, ablation, and stress scenarios
04Out-of-sample evaluation and audit trail
Research lineage
Prior work informing Kernel
Prior work informing Kernel
Building upon open research in financial models, simulation, and collaborative agents.
Kernel's direction is informed by prior work in data-centric financial language models, reproducible market simulation, and agent laboratories that share research artifacts over time.
Data-centric financial LLMs
FinGPT motivates open financial language model infrastructure, automated data curation, and transparent finance-specific model development.
Reproducible market environments
FinRL informs Kernel's emphasis on modular financial environments, historical data, market friction, liquidity, and risk-aware constraints for controlled experimentation.
Collaborative agent research
AgentRxiv provides a useful reference point for shared research memory, report retrieval, and agent workflows that improve by building on prior artifacts.
Applied research infrastructure for analysts, researchers, and institutions working
with complex financial and macroeconomic information.
Macroeconomic intelligence
Agent-assisted analysis of economic releases, central bank communication, fiscal policy, cross-asset reactions, and regime changes.
Financial market research
Systems for studying price dynamics, volatility, liquidity, market structure, risk factors, and changing information conditions.
Institutional research workflows
Multi-agent pipelines for literature review, data gathering, hypothesis testing, model critique, and reproducible research documentation.
Research notes
Selected directions
Selected directions
An index of current research questions, evaluation programs, and systems under study.
Frontier model reasoning over macroeconomic data releases
Multi-agent workflows for financial research and analyst augmentation
Evaluation protocols for AI-generated market hypotheses
Uncertainty quantification in model-driven macro analysis
Benchmark design for agentic financial reasoning systems
Audit trails and source attribution for institutional AI research workflows
Research memory systems for collaborative financial agent workflows
Our team
Researchers and engineers studying AI systems under uncertainty.
Kernel Research brings together quantitative researchers, machine learning specialists, systems engineers, and applied AI researchers building infrastructure for financial and macroeconomic analysis.
The team combines academic depth with implementation discipline: machine learning, high-performance computing, mathematics, and reproducible research infrastructure.
Febin Varghese
Quant LeadEx-McKinsey. MSc Mathematics, First-Class. British Maths Olympiad. Research strategy, mathematical framing, and institutional analysis.
Dr. Yurii Piadyk
Core Systems EngineerPhD NYU in Computer Science. High-performance computing and algorithms. Compilers, solvers, and numerical methods.
Jai Duhra
Lead Developer / Financial Systems EngineeringFull-stack engineer focused on financial systems, data infrastructure, distributed applications, and production-grade research tooling.
Ronan Moynihan
System ArchitectSystems architect and full-stack engineer with experience building market data, execution, and research infrastructure from first principles.
Contact
Research inquiries
Research inquiries
Building AI research systems for finance, macro, or institutional analysis?
Send a short note on the research question, dataset, evaluation protocol, or applied AI workflow. Kernel Research works on exploratory and applied research systems for financial and macroeconomic analysis; it does not provide investment advice, trading services, or live execution systems.