Bynd — AI News Intelligence
News aggregator and summarization pipeline tracking the major AI labs (OpenAI, Anthropic, DeepMind, Microsoft, Meta). Six-step Python pipeline running fully on local LLMs.
By Aditya Singh Khichi, Full Stack Engineer, New Delhi, India.
Tech stack: Python, Ollama, trafilatura, SQLite, RSS.
automated pipeline: 6-step
Problem
Bynd needed to track news coverage of the major AI labs (OpenAI, Anthropic, DeepMind, Microsoft, Meta) at scale. The off-the-shelf options were expensive per-seat news intelligence platforms or hand-rolled Google Alerts that produce noise, not signal. They wanted structured, summarized output they could feed into downstream analysis without paying per-article fees forever.
Approach
Built a 6-step Python pipeline. (1) Collect from Google News RSS, Yahoo Finance RSS, and optional NewsAPI. (2) First-pass classify by company keywords on titles to discard noise cheaply. (3) Resolve Google News redirect URLs and extract article body with trafilatura. (4) Re-classify on full text to catch articles that mention the company in the body but not the title. (5) Generate 30-40 word summaries via Ollama (llama3.2:3b) running locally — no per-token cost, no rate limits, no data leaving the machine. (6) Persist to SQLite plus a CSV export for downstream analysis.
Outcome
A standalone news intelligence pipeline that runs end-to-end on a single machine. Two-pass classification (title-first as a cheap filter, full-text as the recall pass) cut false positives meaningfully without doubling compute. The local-LLM choice traded a small accuracy hit for unlimited throughput at zero marginal cost — the right call for a pipeline that runs every news cycle, not just on-demand.