FUNDA: AI-Native Equity Research Platform | Generated by AI
Question: What is FUNDA (funda.ai)?
Answer: FUNDA is an AI-native equity research platform for institutional public-market investors — basically Bloomberg-terminal-meets-agentic-research-stack, sitting on four layers:
- In-depth Reports (Judgment layer) — analyst-authored deep dives, earnings previews/reviews, thematic coverage. Their marquee case study: a Dec 2024 bullish call on LITE (via Google TPU optical strategy tied to Broadcom/Google/AMD ASIC coverage) is cited as +1474.8% since coverage.
- Research Agents (Intelligence layer) — this is the AI-native part. “Proprietary plays” are prebuilt recurring workflows (e.g. a 573-stock/28-layer semiconductor supply chain sentiment tracker), plus a library of ~12 modular “research skills” (insider Form 4 analysis, 13-F institutional holdings, congressional trading disclosures, options IV/skew, Polymarket earnings-odds ingestion, macro/Fed tracking). There’s also always-on ticker dashboards (e.g. NVDA) that fuse price action, curated news, sentiment, and auto-generated bull/bear “key debate” threads with timestamps and sourcing.
- Live Meetings (Dialogue layer) — scheduled analyst access for institutional clients around earnings/catalysts, with explicit thesis-versioning (“what changed, v2”).
- Expert Database (Evidence layer) — 50,000+ vetted expert network + searchable transcript library from analyst-led calls, compliance-reviewed before being wired back into reports/dashboards as reusable evidence.
Positioning is “analyst-led, expert-backed, AI-native” — not a pure LLM-wrapper stock bot; it’s explicitly trying to combine human analyst judgment + primary evidence (expert calls, filings) + agent workflows into one system, sold to institutional clients (demo-gated, not self-serve retail).
From a build perspective, the interesting technical bits for you: the “Research Agents” section looks like a skills/tools framework not unlike what you’re doing with ww/iclaw/zz — templated agent workflows over a proprietary data layer (13-F, Form 4, options chain, Polymarket) rather than raw LLM chat.
References: