# SentEdge AI

> Boutique AI consulting that ships production-ready systems in weeks, not quarters. Founded by Ryan Price, a certified AI Solutions Architect based in Canada.

## What We Do

SentEdge AI is an AI consulting firm specializing in building real, production-grade AI systems for businesses. We are not a research shop or a vendor of off-the-shelf tools. Every engagement is custom-built, led directly by Ryan Price from requirements through deployment.

Our core value proposition: from requirements to production in 4–8 weeks, with a fixed-scope engagement model, GitHub repo delivery, cloud deployment, and 90 days of post-launch support.

## Consulting Services

We solve 6 categories of hard AI problems:

- **RAG Pipelines & Enterprise Search**: Connect LLMs to your proprietary data using vector databases (FAISS, Pinecone, Weaviate). Your AI speaks with authority about your products, policies, and processes rather than hallucinating.
- **Graph Database Systems (Neo4j)**: When spreadsheets can't capture how data relates, Neo4j reveals hidden patterns. We build knowledge graphs, recommendation engines, and fraud detection systems.
- **LLM Fine-Tuning**: Domain-specific model training with LoRA/PEFT. We've trained BERT and Llama models on custom datasets to achieve production-grade accuracy (40% → 98% documented case study).
- **GPU Optimization & Edge Inference**: NVIDIA Jetson and CUDA-optimized inference pipelines. Low-latency AI at a fraction of typical cloud costs.
- **Real-Time AI Systems**: WebSocket architectures and streaming inference for sub-second response times. No loading spinners.
- **Semantic Search**: Vector embeddings that understand user intent, not just keywords. Users find what they mean.

## Engagement Model

Every project follows 4 phases:

1. **Clarity & Confidence** (1 week): Architecture design, requirements documentation, tech selection. No surprises.
2. **Build** (2–4 weeks): Weekly deliverables, reviewable code, documentation as we go.
3. **Deploy** (1 week): Production environment, security hardening, performance optimization.
4. **Support** (90 days): Bug fixes included, monitoring, knowledge transfer to your team.

## Case Study: AI Chat Assistant

A gaming company's AI support assistant had 40% accuracy and 50–60% hallucination rate with 28–42 second response times. We rebuilt it with an enterprise RAG pipeline:

- FAISS semantic search with 0.75–0.85 similarity thresholds
- Llama 3.2 3B with strict context grounding
- Knowledge base extracted directly from live UI components

**Results**: 98% accuracy, 90% fewer hallucinations, 12–18s response time, 100% grounding rate. Stack: FAISS + sentence-transformers, Flask, CPU-optimized inference (3–4GB RAM).

## Products

- **[Intelligent Knowledge Base](https://sentedge.ai/knowledge-base/)**: Agentic AI system with 3 coordinated agents (Knowledge Capture, Ingestion, Staff-Facing Assistant). Captures tribal knowledge, indexes documents, and gives every employee instant sourced answers. Integrates with Slack, Teams, SharePoint, Google Drive, Salesforce, and more.
- **[x402 Bazaar](https://sentedge.ai/bazaar/)**: Multi-facilitator discovery platform for the x402 blockchain payment protocol. Browse 4,900+ pay-per-use APIs and services across CDP (Coinbase Developer Platform) and PayAI Network facilitators. Designed for both humans and AI agents to make micropayments.
- **[AI Clip Producer](https://sentedge.ai/clip-producer/)**: Autonomous gaming content pipeline. Spectates live matches, records highlights, and generates platform-optimized vertical shorts with AI scene detection, dynamic text overlays, and TTS voiceover. Stack: OpenAI, Remotion, Playwright, FFmpeg.
- **[AI Text-to-Speech Service](https://sentedge.ai/portfolio/)**: Enterprise-grade TTS microservice with word-level timestamp alignment. 4x real-time speed, multi-language, GPU-accelerated, OpenAI API compatible. Built on NVIDIA Jetson.
- **[Replenum](https://replenum.com/)**: Reputation infrastructure for AI agents. Records signed attestations, derives confidence scores from interaction history. Agents register via ERC-8004 or A2A identifiers, settled on Base with USDC.

## Lead Consultant

**Ryan Price** — Lead AI Solutions Architect

- IBM AI Developer Professional Certificate
- Hugging Face AI Agents Course Certificate (2025)
- LinkedIn: https://www.linkedin.com/in/websoftwarecanada/
- Direct client work — no handoffs, no project managers between you and the builder

## Technology Stack

RAG, FAISS, Pinecone, Weaviate, Neo4j, Llama 3.2, BERT, LoRA/PEFT, NVIDIA Jetson, CUDA, Kokoro TTS, Whisper, OpenAI API, Flask, FastAPI, React, TypeScript, Node.js, Remotion, FFmpeg, Playwright, x402 Protocol, Base Network, USDC, Docker, Python

## Contact

Start with our intake form: https://docs.google.com/forms/d/e/1FAIpQLSeN6TJZ-dk9cksjj2Kb5SISGt8RoCKALkOHJHWbZ8DtjE5PYA/viewform

We typically respond within 24 hours. Every project gets a customized proposal.