PoliBerg

About the project

Provide use cases:

Hedge Fund Trader: Gets alerted that Fed rate prediction markets shifted 20 basis points overnight → immediately adjusts bond portfolio positioning before market opens → captures 2% alpha
Prop Desk: PoliBerg detects Polymarket showing 70% odds of new AI regulation → correlates with semiconductor stocks → goes short NVDA before news breaks publicly → 5% profit in 48 hours
Family Office: Monitors geopolitical prediction markets → receives alert that Middle East conflict odds spiked → automatically suggests defensive positioning in gold, oil, and volatility → protects portfolio from 8% drawdown
Crypto Trader: Prediction markets show rising odds of SEC approving Bitcoin ETF → PoliBerg identifies which crypto assets will benefit most → trades before mainstream news → 15% gain in one week

What inspired us Professional desks were manually watching 5+ prediction platforms, copy-pasting into sheets, and still missing most signals. We wanted an intelligence layer that aggregates odds in real time, detects anomalies, and tells you which equities/commodities/crypto/bonds are likely to move.

What we built

Aggregation: unified feed across major prediction markets with a normalized schema.
Intelligence: proprietary models that flag market-moving shifts and correlate them to 10,000+ assets.
Action: alerts, backtesting, and portfolio impact views you can wire into trading stacks via API/WebSocket.

How we built it

News ingestion via Apify (Google News Scraper) to capture headlines/URLs/metadata in near-real time, with deduplication, entity resolution, and clustering before signal detection.
OpenAI-powered symbol linking to fetch related stocks/ETFs from each event/headline by extracting entities and mapping them to tradable tickers.
Airia is our orchestration backbone: one agent links data sources, runs signal & cross-asset mapping, and triggers alerts/backtests/webhooks—secured by RBAC, secrets, audit logs, and policy guardrails.
Data layer that ingests real-time markets (odds, liquidity, order books) and traditional references.
LangGraph-orchestrated agents perform signal detection, cross-asset mapping, and risk scoring.
FastAPI serves sub-second APIs; a streaming pipeline updates dashboards and alert rules.
Backtesting harness quantifies historical alpha from prediction-driven strategies.

What we learned

Prediction markets have hit an inflection point (volume growth, regulatory clarity), making them viable institutional data.
Speed-to-signal matters: odds often shift before terminals and headlines.
Normalization across platforms is essential for robust backtests and real-time alerts.

Challenges we faced

Harmonizing disparate market schemas and resolving event/entity ambiguity.
Avoiding look-ahead bias in backtests and validating true causal relationships.
Balancing sensitivity (catch early moves) vs. noise (false positives) in alerting.
Designing explainable signals that PMs and risk can trust in live workflows.

Who it’s for Quant funds, prop desks, and multi-strategy shops seeking new alt-data alpha; secondarily, platforms that want differentiated feeds for their users.

Why now Volumes and institutional adoption are accelerating, CFTC decisions are clarifying parts of the landscape, and modern AI/infra make large-scale crowd-wisdom analysis feasible. First movers can define the category’s intelligence layer.

Describe the pain of the customer (or the customer's customer):

Professional traders and investment firms are missing massive alpha opportunities because prediction markets reveal market-moving information hours or days before traditional data sources—but there's no way to systematically extract, analyze, or act on these signals.

Outline how the customer addresses the issue today:

Today, sophisticated traders manually monitor 5+ prediction market platforms (Polymarket, Kalshi, Manifold, PredictIt, etc.), spending hours each day:

Copy-pasting data into spreadsheets
Manually correlating prediction shifts with tradeable assets
Missing 90% of relevant signals due to information overload
Paying analysts $150K+/year to track this data inefficiently
By the time they spot the signal, the market has already moved

Show where your product physically sits:

┌─────────────────────────────────────┐
│   Prediction Market Platforms       │
│  Polymarket | Kalshi | Manifold     │
└──────────────┬──────────────────────┘
               │
               ▼
┌─────────────────────────────────────┐
│        PoliBerg Intelligence Layer    │  ◄── WE ARE HERE
│  • Data Aggregation                  │
│  • AI Signal Detection               │
│  • Cross-Asset Correlation           │
│  • Real-Time Alerts                  │
└──────────────┬──────────────────────┘
               │
               ▼
┌─────────────────────────────────────┐
│      Customer Workflows              │
│  Trading Platforms | Portfolios |    │
│  Bloomberg Terminal | TradingView    │
└─────────────────────────────────────┘