AutoSei

AutoSei on Sei EVM uses AI, bots, whale‑tracking & UX for DeFi users by turning complex tasks into simple conversations so they manage portfolios, deploy bots, track whales & optimize yields.

Comment

Inspiration

DeFi is powerful but still in its infancy:

  • Fragmented interfaces force users to juggle wallets, dashboards, and CLIs.
  • Manual strategies miss out on real‑time market moves.
  • Seed phrase risks remain a barrier to mainstream adoption.

We asked: What if complex portfolio management, trading bots, whale‑tracking and yield optimization could be as simple as having a chat? That question drove AutoSei’s vision.

What it does

AutoSei is an AI‑powered DeFi portfolio manager on the Sei EVM network. In one conversational interface you can:

Allocate & Rebalance

AI‑driven weights (w_i) across 7 asset classes by solving:

$$ \max_{{w_i}}\Bigl(\sum_i w_i \mathbb{E}[r_i] - \lambda \sum_{i,j} w_i w_j \sigma_{ij}\Bigr) \quad\text{s.t.}\quad \sum_i w_i = 1 $$

Deploy Trading Bots

Six strategies (Momentum, Mean Reversion, Arbitrage, Grid, DCA, Yield Farming) with ML‑powered entry/exit signals.

Track Whales

Deep‑scan up to 1,000 transactions, set thresholds (\$10 K–\$1 M+), and get AI alerts on potential market‑moving moves.

Optimize Yields

Compare on‑chain protocol APYs in real time and automatically reroute funds for best returns.

Natural‑Language UX

The entire workflow is driven by Google Gemini 2.5 Flash—ask questions like “Rebalance to 20% DeFi, 10% L1, maximize SharpeRatio” and watch it execute.

How we built it

  • Frontend: React 18 + TypeScript + Vite, styled with Tailwind CSS, Shadcn UI & Radix UI
  • Web3 Integration: Wagmi + Viem + Ethers.js for wallet & contract interactions
  • AI Layer: Google Gemini 2.5 Flash via secure API, with dynamic prompt templates for finance
  • On‑Chain Data: SeiTrace API for real‑time whale/tx data; CoinGecko for market feeds
  • Smart Contracts: Solidity contracts on Sei EVM (chain ID 1328), formally verified, with multi‑sig and circuit breakers
  • Deployment: Vercel for frontend; Hardhat & Foundry for contract testing & deployment on Sei testnet

Challenges we ran into

  1. Streaming LLM + On‑Chain Data
    Syncing live blockchain events into async AI prompts without latency spikes.

  2. Security & Compliance
    Ensuring no PII leaks and building multi‑sig/time‑lock protections in contracts.

  3. Backtesting vs. Live
    Bridging historical performance engines with real‑time execution to avoid “paper‑trading” pitfalls.

  4. UX for Complexity
    Designing dialogs that capture advanced financial intents yet remain natural for non‑experts.

Accomplishments that we’re proud of

  • Conversational Portfolio Navigator: Fully functional AI chat demo on testnet.
  • 6 Trading Bots Live: Backtested >500 days of historical data with <10% drawdown.
  • 1,000‑Tx Whale Scanner: Real‑time alerts for “whale” transactions >\$100 K.
  • Automated Rebalancer: On‑chain execution of optimized allocations via smart contracts.
  • Open‑Source Launch: Public repo with docs, contributing guide, and live demo at autosei.vercel.app.

What we learned

  • Prompt Engineering is critical: small tweaks in context drastically change AI’s risk assessments.
  • Latency Matters: batching on‑chain reads with GraphQL reduced data fetch times by 70%.
  • Security First: formal verification and bug‑bounties uncovered edge cases in oracle feeds.
  • User Testing: non‑crypto users loved the chat interface but needed more guided templates.

What’s next for AutoSei

  1. Mainnet Launch & Multi‑chain — extend beyond Sei EVM to Ethereum & Polygon.
  2. Strategy Marketplace — share, rate & copy community‑built bots.
  3. Social Trading & Copy‑Trading — collaborate with top DeFi traders.
  4. Mobile App — on‑the‑go AI portfolio advisor.
  5. DAO Governance — let community vote on feature roadmap and treasury allocations.

Built With

Share this project:

Updates