I'm building an autonomous AI trading agent that makes decisive investment decisions across US equities, Indian markets (NSE/BSE), and Forex. The bot uses Claude Sonnet 4.6 (Anthropic's API) as its reasoning and decision engine — not as a chatbot, but as a structured market analyst that returns BUY / SELL / HOLD decisions with justification.
I am NOT looking for a general AI/LLM developer. I need someone who understands markets AND can build production-grade Python trading infrastructure.
───────────────────────────────────────
WHAT I NEED YOU TO BUILD
───────────────────────────────────────
1. DUAL TRIGGER ARCHITECTURE
- APScheduler (AsyncIOScheduler): market open/close hooks, 15–30 min periodic scans, EOD summary
- WebSocket streams (asyncio): event-driven triggers on price move %, volume spikes, stop-loss/take-profit hits
- Both running simultaneously in a single asyncio event loop
- Per-symbol cooldown timer to rate-limit Claude API calls
2. MULTI-MARKET SESSION MANAGER
- NSE/BSE: 11:45 PM – 6:00 AM ET (via Zerodha Kite API or equivalent)
- London (LSE): 3:00 AM – 11:30 AM ET
- US markets: 9:30 AM – 4:00 PM ET (Alpaca API)
- Forex: 24/5 continuous (OANDA or Alpaca Forex)
- Weekday-only execution with proper market calendar awareness
3. BROKERAGE EXECUTION LAYER
- Alpaca Markets API: US equities + options
- Zerodha Kite or Angel Broking API: NSE/BSE
- Order types: market, limit, bracket (with built-in stop-loss + take-profit)
- Position sizing logic based on portfolio % risk per trade
- Max drawdown kill switch — bot pauses if daily loss exceeds threshold
4. REAL-TIME MARKET DATA
- Alpaca WebSocket: live bars, quotes, trade events for US
- Polygon.io or equivalent: backup data feed
- NSE feed via broker API
- News/sentiment feed (optional): lightweight RSS or Alpaca news stream
5. RISK MANAGEMENT MODULE
- Per-trade risk: configurable % of portfolio (default 1–2%)
- Daily loss limit: hard stop at X% drawdown
- Max open positions: configurable cap
- No pyramiding without explicit Claude confirmation
6. STATE + LOGGING
- Redis: live session state, symbol cooldowns, open position cache
- SQLite or Postgres: full trade log (entry, exit, P&L, Claude reasoning stored)
- Structured JSON logs per session
7. BACKTESTING FRAMEWORK
- At minimum: vectorized backtest on 6+ months of historical data
- Strategy: momentum + mean-reversion hybrid (I will define the logic, you implement the framework)
- Output: Sharpe ratio, max drawdown, win rate, expectancy
───────────────────────────────────────
TECH STACK (NON-NEGOTIABLE)
───────────────────────────────────────
- Python 3.11+
- asyncio + APScheduler (AsyncIOScheduler)
- Alpaca-py SDK (WebSocket + REST)
- Zerodha Kite Connect (or Angel Broking SmartAPI) for India
- Redis (via redis-py)
- SQLite or Postgres
- Docker (containerized deployment on VPS)
- Anthropic Python SDK (claude-sonnet-4-6 model)
NO: LangChain, n8n, AutoGen, CrewAI, no-code tools, or MetaTrader.
───────────────────────────────────────
MUST-HAVES (HARD REQUIREMENTS)
───────────────────────────────────────
✅ You have shipped at least one LIVE algorithmic trading bot connected to a real brokerage (show me)
✅ You understand position sizing, risk-reward ratios, and drawdown management — not just coding
✅ You know asyncio deeply — this is not a synchronous script
✅ Clean, modular, well-commented code — I will maintain and extend this myself
✅ Full code ownership transfers to me — no black boxes, no encrypted modules
✅ You can explain your architecture decisions in plain English
───────────────────────────────────────
NOT WHAT I NEED
───────────────────────────────────────
✗ ChatGPT / RAG chatbot specialists with no trading background
✗ Anyone whose portfolio is only CRM bots, lead gen, or voice agents
✗ MetaTrader EA developers without Python experience
✗ Anyone proposing LangChain or agent frameworks as the core
✗ Copy-paste bots from GitHub — I need custom architecture
───────────────────────────────────────
TO APPLY
───────────────────────────────────────
Start your proposal with the phrase: DUAL TRIGGER
Then answer these three questions:
1. Name one live trading bot you've shipped — what brokerage, what strategy, what language?
2. How would you handle a situation where the WebSocket drops mid-session?
3. What's your approach to preventing the Claude API from being called on every price tick?
Proposals that don't answer all three will not be read.
Apply Now
Apply Now