Fragmented evidence
Trades, charts, macro data, risk assumptions, and research usually live in separate tools.
Independent project · 2026
From a market question to a working system.
QuantFlow is a full-stack quantitative research workspace that connects trade review, risk analysis, macroeconomic context, and original research. I built it to make market decisions easier to examine—not merely easier to record.
01 / PROBLEM
Market analysis loses meaning when its evidence is scattered.
QuantFlow begins with a workflow problem: the information used to make a decision is often separated from the record used to evaluate it later.
Weak feedback loops
A conventional journal records outcomes but rarely connects a decision to its market context and risk process.
Hard-to-audit decisions
When context is scattered, it becomes difficult to test whether a process is consistent, explainable, and repeatable.
02 / ORIGIN
The project grew out of a research question.
My interest in central banks and financial markets led me to study whether Federal Reserve liquidity contains information about equity-market drawdowns. Building QuantFlow became a way to connect that question to an inspectable product.
“Can a market workflow preserve the evidence behind a decision—and turn that evidence into a better research question?”
Design question behind QuantFlow01Study central-bank liquidity and market structure
02Formulate and test the L_net framework
03Translate research needs into product modules
04Build, secure, and test a working full-stack beta
03 / MY WORK
Designed, researched, and developed independently.
I am responsible for the product concept, research connection, interface, analytical logic, application architecture, and deployment. External libraries and public data services are credited as infrastructure—not presented as original work.
Product system
Designed and implemented the responsive interface, navigation, public demo, installable PWA, and connected research workflow.
Trade analytics
Built CSV ingestion, typed trade statistics, equity and P&L analysis, expectancy, streaks, and review surfaces.
Risk and macro
Developed position-sizing tools, explicit risk inputs, macro dashboards, economic context, and FRED data integration.
Full-stack foundation
Implemented authentication, user-scoped PostgreSQL storage, Prisma models, server routes, and fail-closed deployment controls.
04 / RESEARCH
L_net connects the product to an empirical hypothesis.
The working paper tests whether a proxy formed by subtracting Treasury cash and reverse-repo balances from Federal Reserve assets contains information about market downside risk. The measure is an approximation, not a direct measure of liquidity reaching equities.
Lnet = WALCL − TGA − RRP
Focus: asymmetric downside risk
The working paper studies Nasdaq-100 drawdowns with particular attention to left-tail outcomes. Its proposed evaluation includes benchmark comparisons, quantile regression, and drawdown classification. The purpose is to test a falsifiable hypothesis—not to present L_net as a proven trading signal.
View paper and DOIStock-market development in the United States and Russia
My second working paper compares market depth, activity, and institutional conditions in the two countries. It examines how differences in investor access and market institutions relate to stock-market development, while making the limits of a two-country comparison explicit.
View paper, data and DOI05 / ARCHITECTURE
A typed, user-scoped full-stack application.
The architecture keeps interface, analytical logic, identity, data storage, and deployment controls separate. Demo data are synthetic and do not mix with a signed-in user’s workspace.
INTERFACENext.js + ReactResponsive analytical workspace
LOGICTypeScriptTyped analytics and product state
DATAPostgreSQLPrisma and user-scoped records
ACCESSAuth.jsGoogle authentication
DELIVERYVercel + PWACloud deployment and installability
06 / LIMITS
What the current beta does not prove.
Clear limitations matter in both research and engineering. These are the main boundaries of the current project.
Research, not causality
The current L_net work tests predictive relationships in historical data; it does not establish a causal mechanism.
Beta coverage
CSV formats, asset coverage, and model validation remain limited. Wider testing is needed across brokers and market regimes.
No live execution
QuantFlow does not place trades, provide investment advice, or promise that historical analytics will predict future outcomes.
07 / EVIDENCE
Inspect the product, research, and code record.
The demo, both papers, and the source record provide direct evidence of the work and its current stage.
Working productInteractive admissions demoIndependent researchPaper, DOI, and files on ZenodoComparative researchU.S.–Russia paper, data, and DOIEngineering recordQuantFlow source repository
Source access is currently private while production configuration remains in the main repository. A security-reviewed public showcase is the next publication step.