Neelabhro Ghosh@nyx_
2
Events attended
2
Submissions made
India
4+ years of experience
About me
I study computer science and mathematics and prefer building things over talking about them, though I do spend a lot of time talking to AI to test ideas and argue through problems. I work across C++, Java, Python, Go, and Rust, choosing tools based on what actually works. I’m drawn to projects with real constraints, where performance matters and systems need to hold up beyond demos. When something breaks, I dig into it instead of patching it. Lately, I’ve been focused on practical AI systems, especially local and hybrid setups where performance and control matter. I tend to overthink systems, but it usually means they last.
🤝 Top Collaborators
🤓 Latest Submissions
Trade trooper
Trade-Trooper is an autonomous multi-agent options trading engine built on Alpaca. It scans a broad universe of stocks, combining technical indicators, TTM Squeeze momentum analysis, and FinBERT-powered financial news sentiment to identify potential options opportunities. The system separates AI-driven analysis from trade execution. Every proposed trade must pass deterministic mathematical risk gates covering position sizing, portfolio exposure, contract liquidity, drawdown limits, and broker reconciliation before an order can be executed. Trade-Trooper focuses exclusively on defined-risk long Calls and Puts, using automated entry selection and dynamic exit rules based on profit targets, stop losses, holding horizons, and changes in market momentum. Its sector-specific parameters are optimized with Optuna across historical data, allowing different market sectors to use tailored strategy weights. The project integrates Alpaca's Trading API, CLI, and MCP ecosystem, with a Streamlit dashboard for live monitoring, backtesting, hyperparameter tuning, and portfolio management. A local Ollama model generates natural-language explanations and audit summaries, improving transparency without placing an LLM directly in the critical trade execution path. The result is a practical architecture where AI analyzes opportunities, deterministic mathematics governs risk, and Alpaca handles execution.
4 Sep 2026
Vapor Deck
Vapor Deck turns a plain-text topic into a complete presentation without touching PowerPoint or Keynote. You describe what you want to talk about, optionally attach PDFs or reference documents, and the system synthesises your content into a structured narrative before generating each slide as real HTML — with scoped CSS, staggered animations, and live syntax highlighting. Each slide streams token-by-token into an isolated iframe renderer, so you see the layout materialise in real time. A Design AI manages your visual identity: chat with it to set a colour palette, typographic style, and atmospheric feel, and every subsequent slide inherits that brand config automatically. Three built-in themes (dark-tech, clean-light, brutalist) define the base CSS variables the LLM writes against. Once a slide is built, a vision model inspects an offscreen screenshot for layout problems — overflow, clipped content, contrast failures, awkward wrapping — and surfaces a ready-to-paste fix instruction if anything looks off. You stay in control: refine any slide with a natural-language instruction, compare the before and after side-by-side, and export the finished deck as a PDF or standalone HTML files.
10 May 2026
