VOLARIXS
01 / 16
Investor Pitch — Seed Round
Alpha exists.
Find it.
Volarixs is the quant infrastructure layer. Persistent, production-grade ML forecasting — from raw price data to regime-aware, walk-forward validated predictions.
RAISING€2M Seed
STAGEPlatform built, pre-revenue
MARKET~$15.7B quant spend
MODELSaaS — €999/seat/mo
VOLARIXS — CONFIDENTIAL
Daniel Roy, CFA — daniel.roy@volarixs.com
What is volarixs
02 / 16
Test investment ideas in minutes — not dozens of hours with a quant team.
The operating system for financial AI. Pick your universe, goal, and time horizon. Deep Agents assemble the right pipeline from proven building blocks — run it, score it, and keep every result traceable.
Declarative
Paradigms, not hard-coded use cases
Choose universe, target, horizon, and validation. Volarixs builds a repeatable workflow from proven components — so every test is comparable.
Deep Agents
Self-building workflows
Agents assemble the right workflow automatically, apply best-practice validation, and generate a standardized report you can trust and share.
The takeaway
Every asset manager will run ML pipelines. Today, they build them. Tomorrow, they won’t. Volarixs builds and runs the workflow — then explains results in a consistent format.
Research OS — Deep agents
03 / 16
One platform that adapts to each manager’s research style.
Instead of us hard-coding every workflow, you describe the experiment you want. Volarixs turns that into a repeatable process — and applies the same evaluation standards every time.
Declarative research engine
Pipeline compiler
Experiment orchestration layer
L1 — Primitives
Typed building blocks
Features, targets, models, CV schemes, post-processing, evaluation metrics — composable, versioned, and auditable.
L2 — Paradigm DSL
Declarative YAML
A compact spec describing “what” you want to test. Smart defaults get you to a strong first run fast — and you can customize as you learn.
L3 — Composer
Agentic compilation
Deep Agents translate intent to pipelines, enforce guardrails (holdouts, budgets), and produce standardized reports.
Research OS
Volarixs turns “what I want to test” into “here’s the workflow.” As needs evolve, the system adapts — without rebuilding pipelines or waiting on engineering.
The Problem
04 / 16
Every quant team rebuilds the same
infrastructure from scratch.
Pain point 01
Data Pipelines & Feature Stores
Connecting market feeds, computing hundreds of features, maintaining temporal consistency. Rebuilt by every team, every time.
Pain point 02
Model Training & Evaluation
Walk-forward validation, HPO, regime-aware evaluation. Most teams cut corners or get it fundamentally wrong.
Pain point 03
Production Serving & Monitoring
Model selection, prediction pipelines, performance tracking — operational overhead that never ends.
The compounding failure
Nothing
persists.
persists.
When people leave, the infrastructure leaves.
When strategies change, teams start over.
Nothing compounds.
The result: every dollar spent on quant infra is spent again next cycle. Teams never accumulate the institutional edge that the platform should give them.
When strategies change, teams start over.
Nothing compounds.
The result: every dollar spent on quant infra is spent again next cycle. Teams never accumulate the institutional edge that the platform should give them.
6–12
Months to build
$500K+
Before first signal
∅
Persists on exit
The Cost
05 / 16
Billions spent on headcounts and infrastructure, not alpha.
~33K
Quant researchers globally
Hedge funds + asset managers
$15.8B
Total annual HC spend
Fully loaded cost
$473K
Average salary per head
Blended fully loaded
Segment breakdown
| Segment | Headcount | Loaded cost | Total |
|---|---|---|---|
| Quant Hedge Funds | ~18,000 | $563K | $10.1B |
| Quant Asset Managers | ~15,000 | $371K | $5.6B |
Where we capture value
We replace the infrastructure build cost per team. Volarixs eliminates 6–12 months and $500K+ before a single tradeable signal. Even 1% of quant infra spend = $150M+ opportunity.
Spend allocation (estimated)
Infrastructure build
~40%
Alpha research
~22%
Operations & data
~16%
Execution & risk
~12%
TAM → SAM → SOM
TAM — quant spend
$15.7B
SAM — infra budget
$6.3B
SOM — 5yr target
$150M+
The Solution
06 / 16
Volarixs is the quant infrastructure layer.
You describe what you want to test — the platform trains, evaluates, and serves. No quant engineering team required.
Step 01 — Ingest & Compute
Market Data Pipelines
400+ features across 10 families, temporal consistency, multi-asset coverage — handled out of the box.
Step 02 — Train & Evaluate
Model Factory
25+ model types, walk-forward validation, regime detection, Optuna HPO — at factory scale. Zero look-ahead.
Step 03 — Test Hypotheses
Natural Language Interface
Describe what you want to test. Add proprietary data, fund composition, custom targets. No code required.
Step 04 — Predict & Serve
Production Forecasts
Conformal prediction intervals. Regime-aware model selection. Continuously evaluated — nothing stales silently.
The key insight
"Kensho makes data queryable. QuantConnect lets you backtest. Volarixs is the only platform that trains, validates, and serves regime-aware forecasts — built by a derivatives quant who knows where the alpha actually is."
Why Now
07 / 16
Asset managers are under pressure to implement AI/ML — without the teams.
Can't hire
No budget for $500K+ ML engineers.
Can't attract
Top ML talent goes to tech, not finance.
Can't retain
Turnover erases institutional knowledge on exit.
The pressure on buy-side firms
Cut costs — Fee compression, regulatory burden, and margin pressure demand operational efficiency through automation.
Improve returns — Clients demand alpha. Generic tools commoditize returns. Firms need ML-driven signal generation to stay competitive — but the tooling gap is widening as infrastructure complexity grows.
Improve returns — Clients demand alpha. Generic tools commoditize returns. Firms need ML-driven signal generation to stay competitive — but the tooling gap is widening as infrastructure complexity grows.
Built for PMs and analysts. Bought by CIOs.
→
Portfolio Managers
Quant-aware PMs who arbitrate signals. Familiar with factors, risk budgeting — but without a dedicated quant team.
→
Analysts
Enrich fundamental views with quantitative overlay. Test hypotheses, explore regimes without writing a single line of code.
Target segments
01
Asset managers without a robust quant team
02
Discretionary "quant-curious" hedge funds
03
Sophisticated family offices
04
Fundamental HFs seeking quant overlay
Moat: compounding research history
We standardize how research is run — so results compound. Over time, Volarixs builds a growing experiment graph: regime-conditioned performance history, feature+model combinatorics at scale, and reusable workflows that get stronger with every run.
Where We Are
08 / 16
Production-grade ML forecasting, end to end.
Platform built. Pre-revenue. Every core capability is live — the LLM interpretation layer is the only component still in progress.
| Layer | Capability | Status |
|---|---|---|
| DATA | Market data pipelines, multi-asset, temporal consistency | LIVE |
| FEATURES | 400+ features — returns, vol, volume, cross-sectional, cross-domain | LIVE |
| MODELS | LSTM, TCN, MLP, XGBoost, LightGBM, RF, ElasticNet, GARCH family | LIVE |
| VALIDATION | Walk-forward temporal CV, strict out-of-sample, zero look-ahead | LIVE |
| REGIMES | HMM-based regime detection, per-regime model selection | LIVE |
| HPO | Optuna hyperparameter optimization across full model zoo | LIVE |
| UNCERTAINTY | Conformal prediction intervals via MAPIE — calibrated bounds | LIVE |
| PREDICTIONS | Production pipelines with model selection, continuous eval | LIVE |
| LLM | Interpretation layer over full mesh of results | WIP |
Platform pipeline
①
Data Ingestion
Tiingo API integration. Multi-asset coverage. Strict point-in-time data hygiene.
LIVE
②
Feature Engineering
400+ features across 10 families including cross-domain (physics, information theory, astronomy-inspired cycles).
LIVE
③
Model Factory
Prefect-orchestrated training. 25+ model types. Optuna HPO. MAPIE conformal intervals on every output.
LIVE
④
Walk-Forward Validation
Zero in-sample leakage. Rolling re-training. Epoch-level evaluation with aggregated performance recap.
LIVE
⑤
Production Predictions
Regime-aware model selection, continuous evaluation, LLM interpretation layer.
LIVE
⑥
NL Interface (Copilot)
LLM layer over full results mesh. Ask questions, run scenarios, diagnose drawdowns in plain language.
WIP
The Platform
09 / 16
Configure experiments, not code.
Point-and-click experiment design. Select a dataset, choose features, pick a target, select a time window — run. No Python, no Jupyter, no infrastructure setup.
+
Drop screenshot here
or click to browse — the "Experiments → Datasets" UI
The Platform
10 / 16
25 models. Full results exploration.
Every model type is first-class. Full results interface with predictions-vs-actuals, feature importance, residuals, and per-ticker drill-down.
+
ML Engine — 25 models
Drop the model zoo screenshot here
ML Engine — 25 model types
+
Results — predictions vs actuals
Drop the results view screenshot here
Results — predictions vs. actuals
Use Cases
11 / 16
What asset managers actually want to do.
Validated with buy-side portfolio managers. Six core use cases, co-designed with practitioners.
01
Market Regime Analysis
What regime is the market pricing? Layer in inflation, monetary policy, financial conditions. HMM surfaces the hidden state driving returns.
02
Portfolio vs. Cycle
Is the portfolio aligned or fighting the regime? Surface the implied bet, flag mismatches before they become drawdowns.
03
Historical Stress Testing
Replay Liberation Day, the 2022 energy shock, Covid crash. Estimate drawdown and line attribution with full signal decomposition.
04
Shock Analysis
Oil +30%, rates +100bps, VIX to 35. Full transmission map with line-by-line attribution across positions.
05
Thematic Impact
AI disruption, China collapse, deglobalization. Decompose into factors, find historical analogues, estimate forward exposure.
06
Line-by-Line Attribution
Full decomposition by position. Which names drive 80% of the drawdown? Surface the insight — generate the trade idea.
Natural language examples
"What are my remaining risk exposures across sectors given current regime conditions?" · "If rates rise 100bps over 3 months, what are the main risks to my holdings?" · "Test momentum signals on my fund holdings for 5/10/20-day returns with regime-aware models."
Signal Intelligence & Competition
12 / 16
Managed infrastructure, not another toolkit.
We don’t sell signals. We sell the search. Alpha is real, but it’s hidden, regime-dependent, and decays. Volarixs is the industrial search apparatus: factory-scale model generation, regime-aware validation, calibrated uncertainty. You bring the hypothesis. We run the hunt at a scale no small team can match.
Four signal intelligence pillars
01
Regime-aware model selection
HMM selects the right ensemble for current market conditions. Every prediction is conditioned on the detected regime state.
HMM
02
Cross-domain features
400+ features — physics, information theory, astronomy-inspired cycles — applied to financial time series for structural diversification.
400+
03
Calibrated uncertainty
Conformal prediction intervals via MAPIE — real empirical probability bounds, not point forecasts. Position sizing is uncertainty-aware.
MAPIE
04
Factory-scale search
Thousands of model variants across tickers, lookbacks, targets — run automatically, logged, and auditable.
SCALE
Competitive landscape
Kensho (S&P Global)
NLP Q&A over market data
Analytics — backward-looking. Not forecasting infrastructure. No walk-forward validation. No model training.
QuantConnect
Backtesting sandbox
DIY toolkit — you still build everything yourself. No regime detection, no managed infra, no NL interface.
In-house build
Custom ML stack
6–12 months, $500K+, no persistence. Most teams get the walk-forward validation wrong and don't realize it.
Business Model
13 / 16
Platform SaaS — seat-based + enterprise.
Bottoms-up self-serve to land with quant-aware PMs. Top-down enterprise to expand across the institution.
Self-Serve
€999
per seat / month · ~€12K/year
Bottoms-up by quant-aware PMs. Land with the model factory, expand as models go to production. No sales motion required.
→
Individual PM or analyst access
→
Full experiment & signal history
→
Usage-based expansion path
Enterprise
€40K
per year / pilot · expands to €150K–€300K
CIO sales motion. Pilot with proof of value, then team-wide deployment. Enterprise tier unlocks custom data integration and dedicated support.
→
Pilot phase: €40K/yr
→
Team-wide: €150K–€300K/yr
→
Custom data & prop integration
€12K
Self-serve ACV
€150K
Enterprise ACV
>80%
Est. gross margin
SaaS
Recurring revenue
Founder
14 / 16
Daniel Roy, CFA — built by someone who lived the problem.
10+ years in equity derivatives strategy and quant research. Then solo-built the entire platform from scratch during a sabbatical.
2026–NOW
Founder & Solo Engineer
Volarixs
Built production-grade ML platform entirely alone: FastAPI backend, PostgreSQL 6-schema design, Prefect orchestration, 25+ model types, HMM regime engine, LLM interface. Zero external engineers.
2020–2025
Executive Director, Quant Research — Equity Derivatives
JP Morgan
~$40M revenue impact. Systematic signal research on vol surface, equity derivatives, and cross-asset flows. Global team across EU/Asia/NA.
2010–2013
Head of EMEA EQ Derivatives Strategy
Credit Suisse & Newedge
Vol models, systematic research, client-facing strategy. Deep intuition on vol surface dynamics, skew, and derivatives flow signals.
2007
CFA Charterholder
Master in Banking & Finance — Paris II Panthéon-Assas
Prior ventures: Cortexai (computer vision for property, VC fundraising) and Catalyst (e-commerce). E2E product builder with fundraising experience.
First hires — this round funds
MONTH 1–2
GTM Lead
Outbound sales, pilot pipeline, buy-side relationships. First paying customers.
MONTH 2–4
Senior ML Engineer
Production hardening, NL interface, model factory scale-up. Reduces founder single-point-of-failure risk.
MONTH 4–6
Head of Quant Research
Signal expansion, client co-development, buy-side credibility in enterprise sales.
Why this founder
Most quant platforms are built by teams where the finance people can't ship code, the engineers don't understand the market, and the UX looks like it was designed by a quant. Volarixs is built by one operator who's priced the risk, traded the products, and written every layer — including the one users actually see. The platform is the proof.
Use of Funds & Roadmap
15 / 16
Raising €2M — platform built, this round is about clients.
12-month deployment plan
50%
Go-to-market
Customers, pilots, self-serve onboarding, sales, marketing.
€1.0M
40%
Engineering
Production hardening + scale, NL interface shipped to prod, reliability/monitoring, and removing founder single-point-of-failure (2–3 hires).
€800K
10%
Data & Compute
Market data, cloud infra, GPU capacity.
€200K
Targets (next 12 months)
5 paying pilots signed · 2 enterprise contracts · NL interface shipped to production · team scaled to 4 · Series A–ready metrics
Product roadmap — from forecasting to portfolio intelligence
H2 2026
ML Engine — Equities v1
Data — Prop data & fund composition
UX — PM Dashboard
H1 2027
ML Engine — NL hypothesis testing
Data — Alt data integration
UX — Stress testing UI
H2 2027
ML Engine — Multi-asset
Data — Signal marketplace
UX — Self-serve onboarding
Risk — Risk overlay
2028
ML Engine — Meta-learning
UX — Portfolio construction
Risk — Execution integration
VOLARIXS
16 / 16
The Raise
Frontier-level
financial intelligence.
The quant infrastructure layer. From forecasting to full portfolio intelligence — each release opens a new surface for clients. Platform built. Pre-revenue. This round is about scale.
RAISING€2M Seed
USEGTM + Engineering
STAGEPlatform built, pre-revenue
CONTACTdaniel.roy@volarixs.com
The ask
€2M to go from "platform built" to "clients signed". The infrastructure is done. The founder has the market intuition. The raise funds the team and the GTM motion that turns a product into a business.
VOLARIXS — CONFIDENTIAL
Alpha exists. Find it.