Ilyas Rfei@ILY4S
1
Events attended
1
Submissions made
Morocco
2+ years of experience
About me
I am a Software and IoT Engineer with hands-on experience building and deploying AI solutions, particularly under tight deadlines. I regularly participate in AI and data engineering hackathons, where I focus on integrating generative AI workflows into functional applications and bridging the gap between intelligent software and hardware systems.
🤝 Top Collaborators
🤓 Latest Submissions
EarningsEdge
EarningsEdge is an autonomous pre-earnings intelligence platform that transforms a single stock ticker into a full analyst-grade intelligence brief in under 35 seconds. The pipeline runs three data sources in parallel: Bright Data's SERP API and Web Unlocker for live news and full article content, SEC EDGAR for the most recent 10-Q and 10-K filings at authority 1.0, and yFinance for quantitative grounding — consensus EPS estimates, forward P/E, PEG ratio, analyst price targets, and earnings surprise history. What makes EarningsEdge different is not the data — it's what happens to the data. Before a single word is written, a 6-call LLM reasoning chain runs a contradiction audit across all chunks. When a Reuters article says Blackwell production is delayed and the CEO's earnings call transcript says it's sold out for 12 months, the pipeline detects the conflict, resolves it using a strict source authority hierarchy (SEC filing beats transcript beats tier-1 news beats secondary news), and reasons over the contradiction rather than through it. This produces briefs that are internally consistent in ways that single-prompt approaches never achieve. The output has three layers: a raw brief with bull signals, bear signals, risk flags, and resolved contradictions grounded in live sources; a scenario engine with bull, base, and bear cases with confidence scores derived from signal counting rather than LLM gut feel; and a historical pattern matcher that identifies the 2-3 most similar past pre-earnings setups and what happened after. The API serves cached briefs for NVDA, TSLA, and AMD in under 20 milliseconds, with graceful degradation — if live fetch is unavailable, the response includes a transparent data_quality flag rather than silently failing. Live pipeline runs on any ticker in under 37 seconds. Built on FastAPI, LangGraph, Bright Data, SEC EDGAR, and Llama 3.3 70B via AI/ML API.
31 May 2026
