Visit: https://socialvantage.ca

Overview

Social Vantage AI was established in 2026 to solve a persistent gap in social media: reliable, high-signal search across platforms. Instead of searching each network in isolation, the platform aggregates and indexes content from YouTube, X, TikTok, and Instagram to help users identify content hubs, influential creators, and genuinely relevant posts/videos—especially valuable for investors and the advertising industry looking to pinpoint niche opportunities. To deliver high-quality results, Social Vantage AI uses secure, real-time APIs and a grounded datastore that retains user search artifacts (with appropriate security controls). AI models are used to refine messy or ambiguous queries and prioritize the most relevant, high-importance results aligned to the user’s intent. Visit: https://socialvantage.ca

Core Capabilities

• Cross-Platform Search: Unified discovery across YouTube, X, TikTok, and Instagram to surface creators, hubs, and content trends.
• Keyword + Semantic Search: Traditional keyword matching paired with semantic retrieval for intent-based results—plus the ability to expand search to the broader web when needed.
• AI Query Refinement: LLM-assisted query rewriting and ranking to reduce noise and converge on the most meaningful matches.
• Grounded Results Storage: Search results are persisted in a grounded database to support recall, auditing, and consistent user experience.

Analytics, Insights, and Comparisons

Beyond discovery, the platform provides deeper channel and content intelligence—especially for YouTube:

• Channel Statistics: Snapshot analytics and structured channel-level insights.
• Top Videos Explorer: View the top 20 videos for a channel and watch on demand within the platform.
• AI Video Summaries: Generate concise summaries to speed up content evaluation.
• Comment Intelligence: Analyze comments with sentiment analysis and additional NLP metrics to gauge audience reaction and engagement quality.
• Channel Comparison: Compare multiple channels side-by-side and group them into cohorts for benchmarking and pattern detection.

Authentication, History, and User Experience

• Sign-in and Accounts: Users can create accounts using Google OAuth or email authentication.
• Search History Tracking: Authenticated users can track their search history to revisit prior discoveries and iterate on research.
• Client-Side Privacy Controls: Browsing history is stored appropriately in the browser, supporting continuity while maintaining user control.

Technology Stack

• Backend: Python
• Auth: Supabase Authentication (OAuth + email)
• Serverless: Edge Functions (for low-latency workflows and integrations)
• Frontend: React
• Vector Store: MongoDB Vector Store for securely storing semantic embeddings and powering semantic retrieval
• Cloud Runtime: Google Cloud Run
• Container Registry: Google Artifact Registry
• CI/CD: GitHub Actions
• Secrets Management: Google Secret Manager
• Payments: Stripe (handled outside the app for secure checkout and isolation of payment surface area)

DevOps and Deployment

A CI/CD pipeline was implemented using GitHub Actions to standardize releases and reduce operational friction:

• Build & Package: On merge/tag, GitHub Actions builds a Docker container image for the backend service.
• Artifact Publishing: The image is pushed to Google Artifact Registry for versioned, auditable storage.
• Automated Deployments: The workflow deploys the new image to Google Cloud Run, enabling fast rollouts and easy rollbacks.

To keep credentials out of source control and CI logs, the platform uses Google Secret Manager:

• Centralized Secret Storage: Confidential keys are stored securely (no plaintext in repos).
• Runtime Secret Injection: Cloud Run retrieves secrets at runtime for various integrations, including social media APIs and AI model providers.
• Least-Privilege Access: Service accounts are scoped to only the secrets and services required for execution.

Security and Access Control

• OAuth-Based Security: OAuth sign-in with protections in place, including blocking repeated/multiple sign-in patterns.
• Premium Gating (Currently Disabled): The product includes premium conditions for certain AI-heavy features, but premium is currently disabled—the platform is free for all users while validating demand and sponsorship opportunities.
• Secure Data Handling: Semantic data is stored securely in the vector store; search artifacts are retained to improve continuity and grounding.

Challenges and Learnings

The hardest engineering challenge is search robustness: user queries can vary wildly in specificity, language, intent, and ambiguity. That variability increases:

• edge cases (empty intent, overly broad asks, contradictory constraints),
• integration errors (rate limits, API inconsistencies, partial failures), and
• ranking/precision issues (high recall but low signal).

A major focus was building resilient error handling and query refinement so the system can gracefully recover while still producing useful results—without silently failing or returning misleading matches.

Outcome and Traction

Social Vantage AI is currently in active use with hundreds of users globally, including ~50 daily active users, and produces thousands of search results per day. While premium features are planned, the platform remains completely free until it secures sponsors and confirms product-market fit for paid, AI-enhanced capabilities.

This project showcases end-to-end platform delivery: real-time ingestion, semantic retrieval, AI-assisted ranking, analytics, secure authentication, and production-grade CI/CD—built to turn noisy social content into actionable discovery and insight.