.png&w=256&q=75)
🤓 Latest Submissions
Axiom: A trade helper
Axiom is a conversational AI trading agent built entirely on Alpaca's paper trading stack. Instead of a traditional trading UI with a chatbot bolted on, the agent itself is the sole execution path — every trade happens through natural language, with the agent proposing a specific contract and waiting for explicit confirmation before it ever touches the account. Under the hood, Axiom runs on LangGraph connected to Alpaca's official MCP server, giving it real-time quotes, options chains, news, and order execution as native tools — not scraped or simulated data. For its recommendation mode, we trained a separate machine learning model per ticker (AAPL, TSLA, NVDA, and others), using a rigorous, leakage-free methodology: chronological train/test splits, precision-based evaluation instead of misleading accuracy, and a rolling backtest across dozens of historical windows rather than a single lucky split. The model's direction and confidence signal feeds a rules layer that selects the actual options strategy — calls, puts, or defined-risk spreads — depending on both predicted direction and volatility, so the strategy always matches the market condition, not just a price guess. The interface itself, AXIOM Markets, is a dark trading-terminal dashboard showing live candlestick charts, session and fundamentals data, and a persistent agent chat panel where every decision — and its reasoning — is visible before it's ever acted on.
4 Sep 2026
JanSetu — India's Civic AI Assistant
India has over 1,000 government schemes and hundreds of regulatory updates every month — but most citizens and businesses can't navigate them. A farmer in Maharashtra doesn't know he qualifies for PM-KISAN. A startup founder misses a SEBI circular that affects his fundraising. JanSetu is a real-time AI civic assistant that answers any question about Indian government schemes, regulations, and policies — in plain language, in your language. HOW IT WORKS: JanSetu uses a multi-agent LangGraph pipeline with three specialized AI agents: 1. Regulatory Agent — scrapes SEBI, RBI, PIB, GST portals in real-time using Bright Data and Tavily, indexes data into ChromaDB, and answers compliance questions for businesses 2. Schemes Agent — searches myscheme.gov.in, pmkisan.gov.in, startupindia.gov.in and other official portals to match users with eligible government benefits 3. Community Agent — pulls real user experiences from Reddit and Quora so answers include both official information and practical ground-level insights All agents use RAG (Retrieval Augmented Generation) over ChromaDB to ensure answers are grounded in real, current data — not hallucinated from training data. BRIGHT DATA USAGE: Bright Data powers JanSetu's real-time data pipeline. We use the Web Unlocker to scrape JavaScript-heavy government portals like myscheme.gov.in and SEBI's circular listing pages that block standard scrapers. Without Bright Data, these sources are inaccessible programmatically. KEY FEATURES: - Answers any question — from "what is GST?" to "latest RBI repo rate impact on MSMEs" - Supports Hindi, Marathi, Tamil, Telugu and more Indian languages via Sarvam AI - Conversation memory — follow-up questions work naturally without repeating context - Persistent ChromaDB — data is scraped once and cached, giving sub-3 second responses - Confidence scoring — every answer shows a reliability score with source citations - Financial headlines sidebar — live regulatory news feed
31 May 2026
