CH-02 AlgoLend AI · Tanmay Desai

The trading floor

ALGORAND TESTNET

A lending pool on Algorand with three agent desks beside it. Each desk shows what its Python module reads, what it calls, and whether the API actually serves it.

Launch

Not connected yet · algolend.tanmaydesai.xyz the app is being deployed; source is live

The app runs as a seeded demo on testnet: no backend calls, nothing sent on-chain. A live backend sits behind a switch that is off by default.

WIRED

Risk Analyzer

backend/ai/risk_analyzer.py

reads: balance · account age · tx frequency · tx consistency · tx amounts · network activity · reputation
calls: credit_score (0–100) · risk_level (A+ … D) · risk_factors · recommendations

Weights, fixed in code, sum to 1.00

1 balance .25
2 age .20
3 freq .15
4 consistency .15
5 amounts .10
6 network .10
7 reputation .05

SEEDED real `analyze_account` output on synthetic accounts, from `backend/scripts/generate_demo_seed.py`
profile	score	grade
established · 6,200 ALGO, 420 days	86	A+
volatile · 900 ALGO, 120 days	67	B+
new · 15 ALGO, 3 days	54	C+

Served by backend/app.py at /api/analyze-account. A rule score, not a trained model; no labeled outcomes exist to measure it against.

SIMULATED

Market Oracle

backend/ai/market_oracle.py

reads: algo_price · volume_24h · price_change_24h · fear_greed_index
calls: market_sentiment · price_prediction · lending_opportunities · risk_factors · recommended_actions

Imported by app.py, but its market data is generated with random, not read from a feed.

NOT WIRED

Yield Optimizer

backend/ai/yield_optimizer.py

reads: risk_tolerance · investment_amount · time_horizon · 5 pools
calls: optimal_allocation · expected_returns · rebalancing · action_plan

The module exists; app.py doesn't use it.

LendingPool contract

NOT DEPLOYED

depositadds to your balance and the pool total
withdrawpays back the amount plus simple interest since your last update
borrowchecks the pool has the liquidity, then pays out
repayclears your borrowed balance

The rate is stored as interestRate in basis points: 500 bps is 5 %. Interest accrues over Global.latestTimestamp. Written in Algorand TypeScript and kept in the repo, not deployed, so the app simulates deposits and loans.

What was built

Full-stack decentralized lending platform with React, TypeScript and FastAPI (Python), integrating Algorand smart contracts with fast finality and sub-cent fees.
Three AI agent modules for market analysis, risk scoring and yield optimization.