candlestick_chart
Hyde FutureVision Alpha Foundry
Private beta · 60+ strategies in production · 99.97% uptime

Where quant desks turn research into PnL.

Hyde FutureVision is a research-and-execution terminal for crypto perpetual futures, built since 2025 by a quant team who trade this book themselves. Currently onboarding operators by cohort. From hypothesis to live perps in one terminal — no notebook glue, no deploy meetings. The kill switch is a button.

0+
Strategies live
0ms
p50 fill latency
0.00%
90d uptime
ISO 27001
Certified · Sep 2025
Workspace · live
Hyde FutureVision Alpha Foundry — Workspace surface with watchlist, BTC-USDT 15m chart, and running backfill tasks.

Operator screen · production v2.4.0

BTC-USDT-PERP 68,481.25 +1.42% ETH-USDT-PERP 3,825.45 +0.91% SOL-USDT-PERP 161.40 -2.12% XRP-USDT-PERP 0.5283 +0.34% AVAX-USDT-PERP 38.21 -0.61% ARB-USDT-PERP 0.5912 +0.18% DOGE-USDT-PERP 0.1622 0.00% LINK-USDT-PERP 18.42 +2.14% DOT-USDT-PERP 7.31 -0.42%
In production · trailing 90 days

Real numbers from the desk we run.

0+
Strategies live
our desk + design partner
0ms
p50 fill latency
measured operator-side
0.00%
90-day uptime
measured · internal status
ISO 27001
Certified
Sep 2025 · cert on request

Numbers refresh weekly. Cumulative notional is intentionally not on this page. See live status →

Onboarding cohort 01

In private beta with a small group of operators.

We trade our own perps on the terminal every day. Cohort 01 includes our internal desk and one design partner under NDA. New cohorts open quarterly — references available to serious operators after a fit call.

Cohort 01 · active
2 desks
internal + design partner
Cohort 02 · open
Now
accepting applications
On the wire

What ran through the terminal this quarter.

Four operator-attributable moments. PnL figures are conservative — we round to the nearest $1k and stop talking when ops asks us to.

Mar 04 · 09:47 UTC Drawdown guard

The $1.8B perp wick.

Cross-venue basis on BTC ran 22 bps wide for 13 minutes during a coordinated stop-cascade. Our drawdown guard tripped on multiple running strategies within the same minute — flat-and-halt in under two seconds, every time. No human in the loop. Postmortem on file.

$3.1M
drawdown saved · est.
multi-strat
guard tripped
1.8s
slowest flatten
Apr 11 · 14:22 UTC IV signal

FOMC 25 bps cut, signaled 4 minutes early.

The IV surface flagged a regime shift 240 seconds ahead of the press release. Operators on the signal repositioned into the move; their fill quality came in 11.4% better than the post-release average across our book.

+11.4%
fill quality vs peers
11
strats converged
240s
signal lead
May 08 · 23:11 UTC Kill switch

SOL liquidation cascade. Nine strategies flat in 1.4s.

Trade-flow imbalance on SOL crossed the +18σ alarm in under 400ms. One operator hit the kill switch from a phone in a taxi. Nine strategies — three long, six short — flattened on a single ACK. Quote in the field reports below.

$312K
drawdown avoided
9 / 9
strategies halted
1.4s
flatten-to-zero
May 21 · v2.4.0 release Shipped live

Adaptive perf, shipped during a vol spike.

v2.4.0 went out during a live vol regime change. New adaptive performance modes engaged automatically across 487 running strategies — p99 fill latency dropped 14% during the spike itself. No incidents, no rollbacks, no operator pages.

−14%
p99 latency in spike
487
strats live through release
0
operator pages
Inside the terminal

Four surfaces. One operator console.

Every workflow that lives in a trader's day — research, training, deployment, post-mortem — built into one low-latency interface. Live across 14 timezones since November 2024.

candlestick_chart

Workspace

Live candles, indicators, strategies-as-overlays, OHLCV inspector, configurable left watchlist and right task drawer. The screen operators actually leave open all day.

  • 318 instruments tracked
  • 1m · 5m · 15m · 1h · 4h native
  • Layout persisted per operator
science

ML Lab

Train and version brain cores — XGBoost, LightGBM, Transformer, GBM — against the same warehouse that feeds execution. Walkforward-validated, leakage-protected, lineage-tracked.

  • 14,820 model runs to date
  • Median train → deploy: 4h 12m
  • Hash-promotion to StratPak
rocket_launch

Execution Lab

Paper and live runs share the same artifact hash. Diagnostics show feed lags, coverage gaps, engine state. CRITICAL_HALT and KILL_SWITCH are first-class controls.

  • 1.18M fills processed
  • 0 strategies orphaned in prod
  • WS heartbeat ≤ 200ms enforced
menu_book

Trade Journal

Every fill, every bar, every operator decision. PnL in USD + bps, profit factor, max drawdown, per-StratPak attribution, CSV / Parquet export.

  • Tick-level reconciliation
  • 180-day hot retention
  • S3 / Parquet snapshot nightly
From hypothesis to engine

Three steps. No notebooks. No glue code.

01
science

Backtest

Pick a strategy + StratPak, point it at any instrument and timeframe, and get a full equity curve, fill table, and metrics in under 30 seconds. Configurable presets for research vs. production.

timer ≈ 12-30s typical run
02
timeline

Walkforward

Stress-test the same config across rolling windows. The platform tracks fold-by-fold stability so you ship strategies whose Sharpe survives, not strategies that look great in a single window.

verified Leakage-protected, lineage-tracked
03
rocket_launch

Deploy

Promote the validated StratPak into a paper or live run with the same hash. Real-time diagnostics show coverage, slippage, fills. One operator. One tap to halt.

bolt Paper → live on the same artifact
Field reports

Operators on the desk, in their own words.

Anonymized by design — operators talk freely about what works and what doesn't. Real names available under NDA after a fit call.

format_quote Co-founder · internal desk

"The kill switch is a button. Not a chat command, not a config rollback. When the orderbook went wide on May 8 we flattened nine strategies in under two seconds. That's the whole product, honestly."

1.4s
flatten-to-zero
9 / 9
strategies halted
format_quote Design partner · cohort 01

"We replaced four notebooks, a homegrown scheduler, and two Slack channels with one terminal. Walkforward to paper to live is one promotion now, not a deploy meeting."

−61%
research → live time
38 days
in cohort
Why we built it

Because we trade this book ourselves.

Hyde FutureVision started as a quant desk that got tired of stitching notebooks to schedulers to execution scripts. We built the platform we wanted to trade against — and we route our own perps through it every day. The features that survive aren't the ones marketing asked for; they're the ones we couldn't trade without.

folder_data stratpak/alpha_v5.2.toml hash 9942b1a
# Alpha v5.2 — deployed BTC-USDT-PERP @ 15m
[strategy]
name           = "supertrend_rr_v2"
brain_core     = "xgboost"
sharpe_target  = 2.45
config_hash    = "sha256:9942b1a…"

[risk]
daily_loss_cap = 5000      # USD
max_trades_hr  = 150
position_guard = "hard"

[data]
venue          = "primary"
cross_feeds    = ["depth", "trades", "iv_surface"]
ws_max_lag_ms  = 200

[deploy]
mode           = "paper"     # promote via `fv stratpak ship`
operator       = "@gabe"
validation PASSED walkforward 12 folds avg sharpe 2.18 ±0.31 ready_to_ship
Ship cadence

Continuous deployment since launch.

We ship multiple times a week. Every change rides in on a signed commit; every release pins a StratPak compatibility hash. The notable ones below.

v2.4.0 · shipped May 21, 2026
May 21 · 2026
v2.4.0 Major

Adaptive perf during regime change

Engine auto-detects vol regime and pre-scales fanout. Shipped during the May 21 vol spike — p99 fill latency dropped 14% in production during the change itself. Zero operator pages.

May 7 · 2026
v2.3.0 Minor

Fold-stability scorecards

Walkforward now gates promotion by Sharpe percentile across folds — not snapshot performance. Catches "looks great in one window" strats before they bleed in production.

Apr 23 · 2026
v2.2.1 Patch

IV surface as a first-class signal

Options-derived IV surface feeds the ML lab and the execution engine as a per-instrument feature with 1-minute coverage. Operators on it surfaced two basis-arb opportunities in the first 30 days.

Apr 9 · 2026
v2.2.0 Minor

Signed kill-switch audit trail

Every halt now carries a signed operator-ack record with timestamp, user, and post-halt state diff. One customer fail-passed an internal pen-test because the audit trail proved their procedure worked.

Showing 4 of 38 releases. Full changelog arrow_forward
All systems operational

What your seat depends on.

monitoring 90-day window
Execution Engine 100% · 90d

Order routing, fill processing, kill-switch fanout.

Cross-venue feeds 99.94% · 90d

Cross-venue orderbooks, trades, open interest and IV surface.

ML Lab 99.86% · 90d

Training, walkforward, feature warehouse.

Median fill latency · last 24h 11 ms
11ms
p50
28ms
p90
42ms
p99
Last incident
Mar 4 · 2026
9m partial · feeds degraded
Strategies live now
487 running
paper + production
Compliance & security

Passes the same questionnaire your custodian does.

ISO 27001 certified Sep 2025. SOC 2 readiness in progress with Drata. AES-256 + TLS 1.3 in flight, BYOK on Enterprise, audit trail streams to your SIEM.

shield SOC 2 Type II In observation
workspace_premium ISO 27001 Certified · Sep 2025
vpn_key BYOK Per-tenant KMS
key Encryption AES-256 · TLS 1.3
verified Bridge letter On request · NDA
history Audit trail OCSF · 7yr retention
Data coverage

One execution plane. Four data planes. Triangulated against the full tape.

We curate liquidity, depth, open interest, and implied volatility from a curated set of tier-1 venues. Sources are protected under operator NDA — you don't have to broadcast where your alpha is reading from.

Execution
Primary venue
USDT-margined perps · 318 contracts
p50 11ms · p99 42ms
Data plane
Depth + funding
L2 orderbook · cross-venue normalised
WS 8ms · 99.96%
Data plane
Trades + OI
Tick-by-tick · open interest
WS 14ms · 99.91%
Data plane
IV + basis
Options-derived volatility
WS 22ms · 99.88%
v2.4.0 · 487 strategies live

Get a seat on the desk.

We onboard new operators in cohorts. The next intake opens June 9, 2026. Expect a 30-min technical fit call, an SLA, and a credential within 48 hours of acceptance.

Quarterly
Cohort cadence
By hand
We reply to every email
14-day
Paper trial included
lock We don't read your strategy code. Period.