VELQIRIS — Multi-Agent Market Truth & Intelligence Platform

A dual-engine agentic verification and intelligence platform built for autonomous decision-making in acquisitions and algorithmic markets.

1. OpenCV 5 & AWS Serverless Visual Truth Gate (OpenCV AI Competition 2026)

Inspiration & Problem

In autonomous algorithmic trading and visual diligence, multi-agent AI frameworks inspect financial charts (candlesticks, orderflow heatmaps, terminal screenshots). Visual spoofing, tampered high/low wicks, and forged price prints can cause automated agents to execute disastrous capital decisions.

What it does (OpenCV 5 Engine)

• Candlestick Contour Extraction: OpenCV 5 color-space masking (BGR -> HSV) isolates bullish (green) and bearish (red) candle bodies, extracting bounding boxes with cv2.findContours.
• Wick-to-Body Ratio & Anomaly Scoring: Measures vertical upper/lower wicks, detecting manipulated extremes and unphysical candlestick deformations.
• Perceptual Image Integrity Hashing: Cryptographic digests (SHA-256) combined with visual block signatures prevent chart tampering, splicing, and compression artifacts.
• Fail-Closed Validation: Immediately rejects unphysical geometry (e.g. High < Low or inverted body bounds) with HTTP 422 FAIL_CLOSED_VALIDATION_ERROR.

AWS Serverless Architecture

• AWS Lambda: Serverless containerized function executing OpenCV 5 analysis in sub-millisecond compute cycles (lambda_function.py).
• Amazon S3: Immutable evidence vault storing cryptographically hashed verification proofs (template.yaml).
• Amazon API Gateway (HTTP API): REST API endpoint connecting multi-agent frameworks to the visual truth gate.
• Deterministic Test Suite: 9/9 unit and integration tests passing deterministically (test_opencv_market_vision.py).

2. Qloo Taste Graph Cultural Intelligence Agent (Qloo Agentic Hackathon)

Inspiration & Problem

When strategic acquirers evaluate acquisition targets, financial statements only tell part of the story. Cultural alignment, brand fit, and demographic resonance make or break integrations. LLMs typically generate speculative cultural opinions.

What it does (Qloo Engine)

• Qloo Taste Graph Integration: Resolves cultural seeds via Qloo /search and queries /v2/insights to extract affinity correlations across music, literature, dining, and brand domains.
• Fail-Closed Architecture: When API tokens or entity graphs are unavailable, the agent fails closed (QLOO_KEY_NOT_CONFIGURED) rather than fabricating cultural profiles.
• Strategic M&A Diligence: Generates verifiable cultural fit summaries tailored for corporate acquirers and institutional due diligence teams.
• Live Public Demo: Deployed and accessible at https://velqiris-qloo-cultural-fit.netlify.app

How we built it

• OpenCV 5 + AWS Module: Python 3.12, OpenCV 5, NumPy, AWS Lambda, AWS SAM, S3.
• Qloo Agent Module: Node.js serverless functions, Netlify Functions, responsive UI.
• Deterministic Test Verification: 51/51 automated tests passing across the VELQIRIS platform.
• Open Source: MIT licensed.

Accomplishments that we are proud of

• Full OpenCV 5 candlestick computer vision pipeline running serverless on AWS.
• 100% deterministic test coverage across all visual inspection rules and API endpoints.
• Zero mock data contamination: all verifications are mathematically grounded in pixel coordinates, image hashing, or verified API responses.

Built With

• ai-agents
• amazon-web-services
• aws-lambda
• computer-vision
• javascript
• opencv
• opencv-5
• python
• qloo
• serverless
• taste-graph