Built by quants, for bettors
Zaiov Quant AI applies institutional-grade quantitative methods to sports prediction markets — surfacing edge on Kalshi, DraftKings, and FanDuel where the model meaningfully disagrees with the market.
Mission
Prediction markets price outcomes using crowd consensus, which systematically misprices teams outside the public spotlight. We exploit that inefficiency with a calibrated ensemble model trained on thousands of games — not gut feeling, not hot takes. Every pick is gated through our Phase F scoring system, which enforces strict probability, edge, and market confidence thresholds before a signal reaches subscribers.
How it works
ML Ensemble Models
XGBoost + LightGBM + Bradley-Terry with isotonic calibration, retrained weekly via Optuna hyperparameter search.
Quant Scoring
Phase F gate system: convergence, confidence, and market agreement scores. Only picks with QS ≥ 65 earn Quant Select status.
Verified Track Record
Every prediction is timestamped before game time and resolved against official results. Nothing is cherry-picked.
Edge Detection
Model probability vs Kalshi orderbook mid-price. Positive edge with tight spread and strong liquidity = actionable signal.
The Team
Samyam Kafle
CEO & CTO · Co-Founder · ML & Quantitative Systems
Leads the quantitative modeling pipeline — Bradley-Terry ensembles, XGBoost/LightGBM training, isotonic calibration, and the Phase F scoring system. Architected the full prediction infrastructure from data ingestion to Kalshi edge detection.
samyam1kafle@gmail.comChaitanya Vazrala
Co-Founder · Engineering & Infrastructure
Drives backend engineering, server infrastructure, and deployment systems. Oversees the Docker orchestration, CI/CD pipeline, and the data collection architecture that keeps the prediction engine running 24/7.
v.chaitureddy@gmail.com4
Sports covered
NBA · MLB · NCAAM · NCAAWB
1,086+
Picks tracked
Verified & timestamped
83.1%
Backtest QS WR
Phase F qualifying picks
2024–
In operation
Continuous live tracking