# Project: AI-Driven Stock Market System — Signals, **Auto-Trade**, **Scalping**, & **Future Movers** ## Summary Build an end-to-end, **AI-controlled trading stack**: real-time data → AI signals → **auto-execution** with strict **no-loss guardrails** (risk caps & hard stops), plus predictive **stock recommendations** (future high-probability movers). Exchange/broker-agnostic; initial focus **[NSE/BSE + Zerodha/Upstox]** or **[US equities + IBKR/Alpaca]**. > For research/automation enablement. Not financial advice. No performance guarantees; **hard risk limits** and **capital protection rules** are mandatory. ## Core Objectives 1. **AI Algo Suite (with “No-Loss Guardrails”)** * Multi-model ensemble (GBM / Transformer / RL) producing **entry/exit** + **confidence**. * **Capital protection**: hard per-trade stop, trailing stop, per-day max loss, portfolio drawdown cap, **kill-switch**. * **Position sizing** via Kelly-capped or volatility-scaled risk. 2. **Auto-Trade & Execution** * Paper + live modes, **pre-trade checks**, bracket orders (SL + TP), OCO management. * **Latency-aware** routing suitable for **scalping** (see targets below). 3. **Predictive Insights & Recommendations** * **Future risers**: daily & intraday ranking of symbols likely to go up (next session / next 15–60 min). * **Top picks list** with rationale (signals, factors, news/NLP optional). 4. **Risk & Governance** * Exposure caps, circuit filters, time gates, compliance logging, immutable audit trail. 5. **Monitoring & Explainability** * Model drift, feature health, PnL attribution, alerting, dashboards (web). ## Key Features (MVP) * **Data** * Live + historical OHLCV, corporate actions; optional news/NLP and fundamentals. * Resampling (tick/1m/5m/15m/EOD); survivorship-bias-aware universes. * **Modeling** * Signals for: **trend-follow**, **mean-revert**, **breakout-scalp**, **volatility-compression**, **RL policy** (buy/sell/hold/size). * Walk-forward & **purged K-fold** validation; MLflow model registry. * **Scalping Automation** * Micro-targets with **tight SL**, position auto-scale, partial take-profit, time-stop, spread/impact filters. * Venue/broker adapters with **latency budget**; fallback to passive/limit if slippage high. * **Recommendations & Watchlists** * Pre-open and intraday **“Future Movers”** list (top-N symbols with confidence score). * “Why this stock?” tooltips: factors triggered, momentum/volume sweeps, regime tag. * **Execution** * Broker adapters: **[Zerodha/Upstox]** or **[IBKR/Alpaca]** (plug-in architecture). * Order types: market, limit, stop, trailing, bracket (OCO). * **Risk (“No-Loss Guardrails”)** * **Mandatory**: per-trade SL, per-day loss cap, portfolio drawdown cap, cooldowns. * Auto **kill-switch** on breach; manual override; full audit trail. * **Observability** * React/Next.js dashboard: live PnL, positions, heatmaps, drawdowns, order/fill logs. * Alerts to email/Telegram/Slack. ## Nice-to-Have (Phase 2) * Options greeks & spreads; portfolio optimizer. * Alternative data (order book depth, options chain, social sentiment). * Strategy marketplace, multi-account orchestration. ## Tech Stack (Suggested) * **ML/Backend:** Python (FastAPI), Pandas/NumPy, scikit-learn, PyTorch/LightGBM, MLflow. * **Pipelines/MLOps:** Airflow/Prefect, Feast (feature store), Redis, Kafka (optional). * **DB/Storage:** PostgreSQL + TimescaleDB; object storage for artifacts. * **Frontend:** React/Next.js. * **Infra:** Docker, CI/CD, IaC (Terraform), AWS/GCP/Azure. ## Deliverables 1. Architecture + data/contracts + broker adapters. 2. **AI Algo Suite** (signals, scalping module, future-mover ranker) with notebooks. 3. **Execution Engine** (paper + live) with risk guardrails & approvals. 4. **Recommendations UI** (top picks, explanations, confidence). 5. **Risk Policy Configs** (caps, SL templates, cooldown rules). 6. Backtesting + walk-forward reports; paper-trade harness with slippage model. 7. Web dashboard + alerts; complete documentation & runbooks. ## Acceptance Criteria * **Risk guardrails**: SL, max daily loss, and drawdown cap **enforced 100%** (unit & integration tests). * **Scalping latency**: signal→order **≤ 250 ms** target (same-region cloud → broker), or vendor-stated budget. * **Paper vs backtest**: performance within defined tolerance bands (per strategy). * **Recommendations**: daily top-N list with confidence & rationale; offline backtest precision/recall report. * **Reproducibility**: pinned deps, MLflow versions, one-click deploy; immutable audit logs. ## Security & Compliance * Secret vaults, RBAC, encrypted at rest/in transit; broker/regulatory compliance handled broker-side. * Disclaimers throughout; no promises of profit; user-set risk limits required. ## What to Include in Your Bid * Links to 2–3 quant/trading builds (especially with **auto-trade** or **scalping**). * Your **broker/data** plan (APIs, rate limits, fallbacks). * Modeling plan for **future-mover prediction** and **scalping** (features, labels, leakage guards). * Latency profile & infra footprint; monitoring and drift strategy. * Post-deployment support approach. ## Screening Questions 1. Show a **strategy JSON/DSL** expressing: entry from AI signal, SL/TP, time-stop, and size rules. 2. Explain your **walk-forward** & **purged CV** to avoid look-ahead. 3. How do you model **slippage**/fees for **scalping** and trigger **kill-switch** on anomaly? 4. What’s your plan for **future-mover recommendations** (labels, horizon, evaluation metrics)? 5. Provide a **latency budget** and measures to keep signal→order within target. ## Start Your Proposal With `AI-TRADER MVP` — bids without this will be skipped. ## Tags/Skills `Algorithmic Trading` `Auto-Trade` `Scalping` `Machine Learning` `Reinforcement Learning` `Python` `FastAPI` `PyTorch` `Backtesting` `Paper Trading` `Broker API` `Zerodha` `Upstox` `IBKR` `Alpaca` `TimescaleDB` `MLflow` `Airflow` `Docker` `React`