SEDAT

Building AI agents that trade, analyze, and ship.

Scroll
The Operator

A small team. One brain. Full pipeline.

Four specialized AI agents run side by side — a coder that ships, a verifier that breaks, a generalist that fills the gaps, and a quant analyst that walks strategies through rigorous out-of-sample testing. Together they turn a question into a deliverable: a research report, a working script, a backtest PDF, a verified result.

Operative   Sedat
Platform   MiniMax / Mavis
Active since   2026
The Roster

Four agents, four crafts.

Each one is a specialist. Hand off the work that fits their shape.

01 / Generalist
General agent

General

Universal Worker

A flexible operator that adapts to any task. Reads the room, picks the right tool, and hands off to the specialist when the work demands it.

  • Multi-domain task handling
  • Smart delegation to specialists
  • Default front-line responder
02 / Builder
Coder agent

Coder

Software Engineer

Hands-on engineer. Reads code, writes code, ships code. Comfortable across languages and frameworks; turns specs into working systems.

  • Full-stack implementation
  • Refactor & architecture work
  • Bug hunting & shipping fixes
03 / Adversary
Verifier agent

Verifier

Quality Gate

Adversarial review specialist. Tries to break deliverables before they ship — code, research, claims, numbers. The last line of defense.

  • Independent audit of outputs
  • Edge case & assumption attack
  • Pre-ship sign-off
04 / Analyst
Backtester agent

Backtester

Quantitative Analyst

Pulls years of price data, runs strategies through walk-forward validation, kills the brittle ones, and outputs a BUY / SELL / HOLD / AVOID verdict with a full PDF report.

  • Regime-aware strategy selection
  • 9-gate robustness filter
  • Concrete trade plans + PDFs
Featured — Backtester v2
Backtester pipeline
Featured Agent

A quant desk in a single agent.

Backtester doesn't just run a single backtest. It walks a strategy through rolling out-of-sample windows, applies nine robustness gates — Monte Carlo, parameter sensitivity, sub-period consistency, correlation dedup — and emits a verdict, a confidence level, and a concrete trade plan you can actually execute.

P1 / 01
Asset & Regime Profiling
Pulls adjusted OHLCV, checks liquidity, classifies the current regime (trending / ranging / volatile) to pre-select matching strategies.
P2 / 02
Walk-Forward Validation
Rolling 2-year train / 6-month test windows. Optimizes on train, freezes parameters, evaluates on untouched test. No lookahead bias.
P3 / 03
Robustness Gates
9 strict filters — profit factor, max drawdown, sub-period consistency, Monte Carlo, parameter sensitivity, regime survival, correlation dedup.
P4 / 04
Sharpe-Weighted Ensemble
Survivors vote, weighted by their real out-of-sample Sharpe. A Bull Score and Bear Score roll up to a single composite bias.
P5 / 05
Verdict & Trade Plan
Outputs BUY / SELL / HOLD / AVOID with confidence. Computes entry zone, ATR-based stop loss, 1:2 take profit, exact position size from your capital and risk.
P6 / 06
10-Section PDF
Cover, executive summary, regime, strategy table, ensemble, trade plan, equity curve, monthly heatmap, risk analysis, disclaimer. Ready to share.
By the Numbers
0 Active Agents
0 Robustness Gates
15+ Strategies in Library
10 Section PDF Reports
The Stack

What runs under the hood.

yfinanceMarket Data
ccxtCrypto Data
vectorbtBacktest Engine
backtraderEvent Engine
pandas / numpyAnalysis
scipy / optunaOptimization
matplotlibCharts
reportlabPDF Reports
Next Move

Spin up your own team.

Drop a ticker. Get a verdict. Or build a custom agent that does whatever you need it to do.

Meet the Agents See Backtester in Action