Unfluke AI - Your Trading Assistant

• Company Fundamentals - Q&A on any NSE-listed name. Ratios, history, peer comparison, balance sheet flags. Replaces 15 minutes of clicking through Screener and annual reports with one question.
• Fundamental Screener - describe a screener in English, get back a runnable filter. "Midcaps with ROE > 18%, P/E < 25, debt < 0.5" instead of clicking 8 dropdowns.
• Scanner - same idea for technicals. "Stocks where RSI > 70 on daily and close > VWAP on 5-min."
• Basic Backtest - describe an option strategy and get a runnable backtest spec. "Bank Nifty short straddle, 9:20 entry, 30% combined SL" → ready to test.
• Advanced Backtest - multi-indicator equity and crypto strategies with risk management, expressed in English, translated into expression-based backtest rules.
• YouTube Strategy - paste a strategy-video URL, the bot watches it, extracts the rules, and gives you a runnable backtest. Sounds gimmicky until you try backtesting a YouTube strategy by hand.

• "Walk me through RELIANCE's last 5 quarters of cashflow"
• "Compare TCS vs INFY on ROE, debt, and growth"
• "Suggest an option strategy for a mildly bullish view on BANK NIFTY"
• "Why did HDFC BANK drop 6% this week?"
• "Find midcaps with ROE > 20%, P/E < 30, sales growth > 15%"
• "Test a Bank Nifty Thursday short straddle from 2022 with 30% SL"
• "Build a 3-leg iron fly on Nifty with 100-point wings"
• "What does Bollinger Band squeeze mean and when does it matter?"

We are explicit about what Unfluke AI is not:

• It does not predict prices. No "buy tomorrow" signals. No "AI says short this stock". Markets are an adversarial system and persistent prediction accuracy is vanishingly rare; any product that markets otherwise is selling fiction.
• It does not generate trade signals. The bots help you build research and strategies. Whether to trade them is your decision, every time.
• It does not place orders. Execution stays with your broker. We're broker-agnostic.

Read the broader piece on AI in Trading for why those boundaries matter.

Powered by current-generation large language models on dedicated infrastructure, with a vector store that retrieves Unfluke's curated knowledge base - disclosures, annual reports, regulatory filings, the indicator library, and historical market data - before generating an answer. The technical term is "retrieval-augmented generation" (RAG). The practical effect: fewer hallucinations, more grounded reasoning, and answers that cite actual Indian-market data rather than guessing from training data that may be months stale.

Four front doors, same brain:

• In-app - at /in/chatbot
• Telegram - @UnflukeAI_bot
• WhatsApp - same account, broker-style conversations
• Twitter mention - ping us for public Q&A

Curious before signing up? The chatbot has a demo mode for unauthenticated visitors (IP-tracked, limited daily messages). Try a few queries to feel what each bot type does. Full access and higher quotas come with free signup and paid plans respectively.

The mental model that produces the most value from Unfluke AI: treat it as a research intern who reads fast, drafts well, and occasionally gets things subtly wrong. Use it for first drafts of research, scanner specs, and backtest setups. Always verify the output before committing. The intern model captures both the value (huge time savings) and the limits (don't outsource judgment) of AI tools in 2026.

Most general-purpose LLMs are trained primarily on US-market content. They miss Indian-market nuance: NSE-specific instruments, F&O conventions, SEBI regulation, Indian tax structure, broker terminology, and Hindi-language financial commentary. Unfluke AI is grounded in Indian-market data and tuned for the Indian retail trader's workflow. The difference shows up the first time you ask about a Mid Cap Nifty constituent or a sector-specific F&O setup.

These are ideas to study, paper-trade and backtest, not recommendations or signals, and not a promise of any result. The assistant can be wrong; check what it gives you before you rely on it. Size, entries, exits and risk are your own decisions.

1. Let it draft, then test it yourself

• Setup: describe a strategy in plain English, for example a short straddle with a combined stop-loss, and ask for a backtest.
• Use: read the rules it produced, correct anything that differs from what you meant, then run the test.

2. Turn a screen idea into a filter

• Setup: ask for midcaps with ROE above 20%, P/E below 30 and sales growth above 15%.
• Use: a starting shortlist that you then check on the fundamentals pages.

3. Compare two companies quickly

• Setup: ask it to compare two names on ROE, debt and growth.
• Use: a quick first read; confirm the numbers on the company's own pages.

4. Test a video strategy before trusting it

• Setup: paste a strategy video link into the YouTube Strategy bot.
• Invalidation: if the backtest of the extracted rules does not hold up, the video's claims don't either.

How to test these

Take whatever the assistant drafts, run it on Unfluke's backtester or scanner, and judge the result yourself.

AI compresses time and surfaces ideas. It doesn't replace judgment. Every AI output is a draft - verify the data point, review the scanner spec, sanity-check the backtest before you commit capital. The AI also occasionally gets factual details wrong, particularly on rapidly-changing data; cross-check anything critical against the actual source. The AI does not predict markets reliably and never will.

Sign up free for 5 AI messages a month, or read the broader pieces: AI in Trading, AI Trading Software, and AI Stock Market Analysis.