DipScalar, Kaan Kont, Parshawn Gerafian, Tejveer Bagri, Ben Shamloufard

https://dipscalar-production.up.railway.app/

Cross-platform batch trading terminal combining Polymarket prediction markets with Liquid perpetual futures.

DipScalar is an AI-powered trading interface that lets users search market themes in plain English and instantly discover relevant opportunities across Polymarket prediction markets and Liquid perpetual futures in one place. Instead of manually scanning separate platforms, users can type queries like "show me Venezuela markets," "hedge against inflation," or "crypto bull run" and get back a structured, themed set of positions spanning both venues.

Built for the Liquid Trading ($8K) and Polymarket Bonus ($2K) hackathon tracks.

Inspiration

Trading across fragmented platforms is slow, repetitive, and mentally expensive. A user may have an idea or macro view — for example, rising inflation, geopolitical instability, or election uncertainty — but acting on that thesis often requires manually hopping between venues, searching for related markets, figuring out which instruments are relevant, and then placing each trade separately.

We wanted to build a system that feels closer to how people actually think. People do not think in tickers first — they think in themes, narratives, events, and risks. DipScalar was built around that idea. Rather than forcing users to translate a thought into individual markets themselves, we use AI to interpret the theme, search both prediction markets and perpetual futures simultaneously, and present the results as a unified batch that can be acted on quickly.

The core inspiration was to create a trading workflow that is more natural, more thematic, and more powerful than the current platform-by-platform experience.

What it does

DipScalar allows users to search for a market idea in natural language and see related opportunities across two different trading ecosystems at once:

Polymarket, for event-driven prediction markets
Liquid perpetual futures, for directional exposure through tradeable perps

A user can type something like:

"show me Venezuela markets"
"hedge against inflation"
"oil markets"
"crypto bull run"

The AI agent then interprets the request, generates relevant search terms, identifies matching Polymarket events, resolves related Liquid instruments, and returns the results in a clean grouped interface.

The output is organized into collapsible event groups for prediction markets and individual perp cards for Liquid instruments. Users can inspect the results, view mini price charts, configure positions such as Yes / No or Long / Short, choose trade sizes, and build a thematic batch from one unified search flow.

The product is designed to make thematic trading much faster. Instead of treating prediction markets and futures as separate workflows, DipScalar treats them as part of the same decision-making environment.

How we built it

DipScalar is built as a full-stack application with a React frontend and a FastAPI backend, with an AI orchestration layer connecting user intent to real market discovery.

AI query interpretation

At the center of the system is an AI agent powered by Claude Haiku. When a user types a natural language prompt, the model parses the request and produces three things:

the likely user intent
a set of relevant search terms
a set of candidate Liquid symbols related to the theme

This lets us move from vague thematic prompts to structured multi-platform market discovery. For example, a prompt about Venezuela might lead to search terms involving leadership, oil, sanctions, elections, or geopolitical instability.

We also spent a significant amount of time refining the searching and retrieval algorithm so that it performs well on cheaper models while still feeling quick and responsive. That was an important design goal for us: we did not want the product to depend on an overly expensive model stack just to feel useful. Instead, we focused on making search quality, symbol matching, query expansion, and deduplication strong enough that the system remains cost-efficient, relatively fast, and practical to use in a real product setting.

Polymarket integration

For prediction market discovery, we integrated with the Polymarket Gamma API and used server-side search rather than trying to brute-force large event lists client-side. This was important because Polymarket has a large volume of events, and a simple local filtering approach would miss many relevant results.

We group returned prediction markets by event so users are not overwhelmed by dozens