SEMICONDUCTORS — AI News Platform

AI-powered semiconductor news and analytics platform built on top of the PUMP trading engine, combining price charts with curated articles, LLM insights, and chat for a single institutional client.
Nuxt 3FastAPIPrefectPythonLLMClient Project
2024 - 2025

Overview

SEMICONDUCTORS is an AI-powered news and analytics platform focused on the semiconductor industry. It combines price and signal charts with curated articles, LLM-generated insights, and an embedded chat experience, built as a solution for a single large institutional client.

The product is essentially a rebirth of the PUMP project: it reuses the trading and charting foundations from PUMP, but pivots the experience toward news and research workflows rather than social-media feeds.


Relationship to PUMP

  • Reused the proven PUMP infrastructure for time-series data, charting, and sentiment overlays.
  • Added a dedicated news layer: articles, summaries, and AI insights tied to tickers and semiconductor themes.
  • Extended the UI to surface:
    • multi-panel chart layouts
    • AI-generated article insights
    • a ticker-aware chat panel (PUMP Chat) for research questions.

The result is a focused platform where analysts can move seamlessly between performance charts, relevant news, and LLM assistance in a single screen.


My Involvement & Timeline

  • Jul 2024 – Aug 2024 – First phase
    • Forked and adapted PUMP into the SEMICONDUCTORS product.
    • Wired the initial Nuxt 3 SPA interface to FastAPI endpoints.
    • Implemented layouts for charts, article feeds, and AI insights.
  • Sep 2024 – Feb 2025 – Project on hold
    • Core work paused while I focused on the QSIM project.
  • Feb 2025 – ~May 2025 – Second phase
    • Returned to SEMICONDUCTORS to deepen the AI news experience:
      • Built daily and weekly summaries on a dedicated page.
      • Added parsers for new news sources into the pipeline.
      • Refined the UI/UX for insights, articles, and chat.

Key Product Features

AI News & Insights

  • Aggregates semiconductor-related news from multiple sources.
  • Uses LLMs to:
    • summarize long-form articles,
    • extract key drivers, risks, and sentiment,
    • surface ticker- and theme-specific insights.
  • Displays insights alongside charts, so users can connect price moves with the news context behind them.

Daily & Weekly Summary Pages

  • Dedicated page for daily and weekly semiconductor market digests.
  • Summaries generated by LLMs from the latest articles and signals.
  • Structured sections such as:
    • key movers and themes,
    • notable company-specific events,
    • macro/industry-level developments.
  • Designed to be a quick “morning briefing” for the client’s research team.

Chat & Research Workflow

  • Embedded chat panel that understands tickers and topics (e.g., “Trade ideas about NVDA?”).
  • Chat responses blend:
    • knowledge from recent news,
    • general company context,
    • links back to original sources when needed.
  • Conversation history kept within the current asset or view to maintain context.

Data & Orchestration

  • Implemented and extended Python + Prefect flows for:
    • fetching articles from multiple news providers,
    • normalizing and cleaning text,
    • routing content to LLMs for summarization and tagging,
    • storing processed data for charts, insights, and summaries.
  • Added new parsers during the second phase to broaden coverage and improve signal quality.
  • Coordinated Prefect schedules, retries, and monitoring so daily and weekly summaries arrived reliably for the client’s workflows.

Tech Stack

  • Backend
    • Python
    • FastAPI (REST APIs for charts, news, insights, and chat)
    • Prefect (ETL and orchestration for news ingestion and LLM pipelines)
  • Frontend
    • Nuxt 3 (SPA mode)
    • Vue 3 Composition API
    • Vuetify as the main design system for layouts, navigation, and cards

SEMICONDUCTORS showcases how the original PUMP trading engine can be repurposed into a news-centric research product, layering LLM-powered summaries and chat on top of familiar analytics views for a single high-value client.

Copyright © 2026