Case study · (c)
Market Pulse Watch
Seismopulse · market anomaly & risk desk
1 page
Candace Hagood
Product & Quality Leader
Outcome
Unusual market moves surface early, with a reason and a track record.
- ✓One desk replaces checking six separate sources
- ✓Alerts arrive ranked by severity against the user’s own thresholds
- ✓Every AI pick is graded daily, so trust is earned, not assumed
Problem
Unusual market activity — sudden volume, large block trades, earnings surprises, new contracts, geopolitical or cyber events — is spread across many sources and easy to miss until it's too late.
My role · focus · quality · key features›
My role
Product owner and designer. I chose the signals and data sources, designed the desk layout, the risk index and alert rules, and defined how the AI explains — and is scored on — its recommendations.
Product focus
A single, dense desk that surfaces what's unusual right now, explains why, tracks whether its own calls were right, and lets the user arrange it their way.
Quality & validation
Independent data feeds compared side by side. A playbook records every recommendation with its lineage and tracks daily accuracy, so the AI is accountable. Recommendations are framed as information, not advice.
Key features
- ·Spike, volume and late-day anomaly detection across NASDAQ and global tickers
- ·Block trades, open bids and earnings tracking
- ·IPO calendar and government contract news with detail views
- ·Global threat index (conflicts, country risk, exploited CVEs, malware) mapped to affected companies and sectors
- ·Reddit social sentiment and news feed
- ·AI playbook with recommendation lineage and accuracy tracking
- ·Holdings, custom alert rules, thresholds and sound alerts
- ·Market schedule and drag-to-reorder panels
▤ README.md›
# Market Pulse Watch > Seismopulse · market anomaly & risk desk "Seismopulse": a real-time anomaly desk that combines volume spikes, block trades, earnings, IPOs, government contracts and global threat intelligence into one explained market read. ## Outcome **Unusual market moves surface early, with a reason and a track record.** - One desk replaces checking six separate sources - Alerts arrive ranked by severity against the user’s own thresholds - Every AI pick is graded daily, so trust is earned, not assumed ## Problem Unusual market activity — sudden volume, large block trades, earnings surprises, new contracts, geopolitical or cyber events — is spread across many sources and easy to miss until it's too late. ## Solution Seismopulse gathers price, volume, earnings, IPO, contract, news and threat signals into one desk, flags what's unusual against your thresholds, and explains each call. Every recommendation is then tracked day by day, so the AI is held to its own record. | Problem | Solution | | --- | --- | | Unusual moves are missed until it's too late. | Spike, volume and late-day anomaly detection ranks alerts by severity, with custom alert rules. | | Contracts, news and geopolitics live in different places. | Contract awards, news and Reddit stories are scored for impact and mapped to the stocks they could move. | | AI picks can't be trusted without a track record. | Each recommendation is tracked daily as hit, partial or miss, with its evidence and lineage. | ## Features - Spike, volume and late-day anomaly detection across NASDAQ and global tickers - Block trades, open bids and earnings tracking - IPO calendar and government contract news with detail views - Global threat index (conflicts, country risk, exploited CVEs, malware) mapped to affected companies and sectors - Reddit social sentiment and news feed - AI playbook with recommendation lineage and accuracy tracking - Holdings, custom alert rules, thresholds and sound alerts - Market schedule and drag-to-reorder panels ## Tech stack - TanStack Start (React SSR) - TypeScript - Tailwind / Radix UI - Lovable Cloud (Postgres) - Lovable AI Gateway (Gemini) - Yahoo Finance, Finnhub - USAspending.gov - NIST NVD - Reddit, Google News - Recharts ## Notable components - Self-scoring AI playbook with day-by-day hit/miss tracking - Global threat index (conflict, country risk, exploited CVEs) mapped to tickers - Contract-bid news scored for stock impact and reach ## Scope - 6 live data sources combined in one desk - Every AI recommendation scored daily against actual price moves ## Links - Live: https://anomaly-guard-now.lovable.app ## Author Candace Hagood · Product & Quality Leader · AI Engineering
Skills demonstrated
Data-heavy UISignal designExplainable & accountable AIInformation hierarchy
How the solution answers the problem
Seismopulse gathers price, volume, earnings, IPO, contract, news and threat signals into one desk, flags what's unusual against your thresholds, and explains each call. Every recommendation is then tracked day by day, so the AI is held to its own record.
- ProblemUnusual moves are missed until it's too late.SolutionSpike, volume and late-day anomaly detection ranks alerts by severity, with custom alert rules.
- ProblemContracts, news and geopolitics live in different places.SolutionContract awards, news and Reddit stories are scored for impact and mapped to the stocks they could move.
- ProblemAI picks can't be trusted without a track record.SolutionEach recommendation is tracked daily as hit, partial or miss, with its evidence and lineage.
Measurable scope
- ·6 live data sources combined in one desk
- ·Every AI recommendation scored daily against actual price moves
Notable & novel components
- ·Self-scoring AI playbook with day-by-day hit/miss tracking
- ·Global threat index (conflict, country risk, exploited CVEs) mapped to tickers
- ·Contract-bid news scored for stock impact and reach
Tech stack
TanStack Start (React SSR)TypeScriptTailwind / Radix UILovable Cloud (Postgres)Lovable AI Gateway (Gemini)Yahoo Finance, FinnhubUSAspending.govNIST NVDReddit, Google NewsRecharts
Why it matters
Shows I can design for power users who need a lot of information — and build in the checks that make an AI's output trustworthy.
How it works · From raw signals to an accountable call
- 01
Ingest
Prices, volume, block trades, earnings, IPOs, contracts
- 02
Enrich
Reddit sentiment, news, global threat index
- 03
Detect
Spike, volume and late-day anomalies vs. thresholds
- 04
Explain
AI recommendation with the drivers behind it
- 05
Score
Playbook tracks lineage and daily accuracy
Threat sourcesConflict zonesCountry riskExploited CVEsMalware telemetry
Inside the app
Open live product ↗anomaly-guard-now.lovable.app