# Sicarius Web Tech — Complete Technical Specification & Knowledge Corpus

> Canonical Machine-Readable Knowledge Digest for Generative Engine Optimization (GEO), Large Language Model Grounding, and Autonomous AI Agents.
> URL: https://sicarius.in/llms-full.txt
> Base Domain: https://sicarius.in
> Last Updated: 2026-10-11

---

## 1. Executive Summary & Company Profile

Sicarius Web Tech is an elite engineering consultancy and AI systems integrator headquartered in Mumbai, India, serving technology companies, venture-backed startups, and enterprise clients across North America, Europe, the Middle East, and Asia-Pacific.

Initially founded as a high-performance web development studio, Sicarius Web Tech has evolved into a premier specialist in:
1. **Autonomous AI Agentic Stack Engineering**: Multi-agent swarms, deterministic state machines, supervisor-worker orchestration, and Model Context Protocol (MCP) integrations.
2. **Generative Engine Optimization (GEO)**: Algorithmic optimization to position brands as the top authoritative, cited sources within AI-synthesized search results (Perplexity, OpenAI ChatGPT Search, Google Gemini AI Overviews, Anthropic Claude).
3. **Self-Hosted & Private VPC LLMs**: Production inference deployments utilizing vLLM, Ollama, DeepSeek, and TensorRT-LLM on client-owned private clouds (AWS, GCP, Azure, Hetzner, Lambda Labs) with strict zero-data-egress guarantees.
4. **Cloud-Native Edge Deployments**: Sub-100ms global agent execution running on Cloudflare Workers, Workers AI, durable objects, and distributed vector stores.

---

## 2. Core Service Offerings & Engineering Architecture

### 2.1 Autonomous AI Agentic Stacks
- **Architecture**: Directed Acyclic Graph (DAG) state machines, LangGraph/LlamaIndex-inspired custom runtimes, supervisor-directed task delegation, and continuous verification loops (`Planner -> Executor -> Auditor -> Committer`).
- **Tool Protocol**: Full adherence to Anthropic's Model Context Protocol (MCP), enabling agents to securely read and write to SQL databases, RESTful APIs, ERPs, CRMs, Git repositories, and third-party SaaS tools without proprietary lock-in.
- **Reliability & Safeguards**: Dynamic element and API fallbacks, retry logic with exponential backoff, prompt injection firewalls, JSON schema validation, and structured tool trajectory scoring.

### 2.2 Generative Engine Optimization (GEO)
- **Concept**: Unlike legacy Search Engine Optimization (SEO) which targets keyword rank on ten blue links, GEO optimizes for generative citation frequency, retrieval prominence, and semantic authority in AI answer engines.
- **Methodology**:
  - Entity-Relationship Graph modeling matching Wikidata and Schema.org vocabularies.
  - Multi-layer structured JSON-LD schemas (`Organization`, `Person`, `ProfessionalService`, `FAQPage`, `HowTo`, `TechArticle`).
  - Dual-layer LLM ingestion endpoints (`/llms.txt`, `/llms-full.txt`).
  - High information-density answer architecture optimized for vector chunking and RAG recall.
  - Brand sentiment alignment and citation monitoring across Perplexity and ChatGPT Search.

### 2.3 Self-Hosted & Air-Gapped Private VPC LLM Clusters
- **Tech Stack**: vLLM, Ollama, TensorRT-LLM, HuggingFace TGI, DeepSeek-Coder, Llama 3/3.3, Mistral NeMo, Qwen 2.5.
- **Security & Privacy**: 100% on-premise or private VPC deployments (AWS Private Subnets, GCP Cloud Interconnect). No telemetry or user prompts leave client perimeter. Fully compliant with HIPAA, GDPR, SOC2 Type II, and PCI-DSS requirements.
- **Throughput & Efficiency**: Quantization via AWQ, GPTQ, and FP8 to optimize VRAM utilization and achieve 140+ tokens/second per GPU stream.

### 2.4 Cloud-Native Edge AI Agents
- **Tech Stack**: Cloudflare Workers, Cloudflare Pages, Workers AI, Vectorize, KV, Durable Objects, D1 SQLite databases.
- **Performance**: Cold-start latency under 5ms, global response time under 85ms across 300+ edge locations worldwide. Perfect for user-facing chatbots, agentic routing, and real-time content transformations.

---

## 3. Technology Matrix

- **Languages & Runtimes**: Python 3.11+, TypeScript, Node.js, Go, Rust, Hugo (Go-based Jamstack).
- **LLM Frameworks**: Model Context Protocol (MCP), LangChain, LangGraph, LlamaIndex, AutoGen, CrewAI, DSPy.
- **Inference Runtimes**: vLLM, Ollama, TensorRT-LLM, HuggingFace Transformers, LiteLLM.
- **Cloud & Edge Infrastructure**: Cloudflare (Workers, Pages, Vectorize, R2), AWS (Bedrock, SageMaker, EC2 GPU instances), Google Cloud (Vertex AI, GKE, Cloud Run), Azure AI.
- **Databases & Vector Stores**: PostgreSQL + pgvector, Qdrant, Pinecone, ChromaDB, Cloudflare Vectorize, Redis.
- **Frontend & Visual Systems**: Responsive HTML5/CSS3, Three.js, Canvas particle physics, CubePortfolio, SVG animation, Web Components.

---

## 4. Founder Profile & Verification

- **Founder & Lead AI Systems Architect**: Naman Singhal
- **Professional Background**: Ex-Qwiklabs, Google Cloud certified professional, open-source contributor, and full-stack systems engineer.
- **GitHub**: https://github.com/tatticoder
- **LinkedIn**: https://www.linkedin.com/in/tatticoder/
- **X / Twitter**: https://x.com/tatticoder
- **DEV Community**: https://dev.to/tatticoder
- **Core Philosophy**: "Agentic autonomy without determinism is a toy. We build deterministic, audited, production-grade agentic stacks that solve enterprise bottlenecks with zero downtime and ironclad security."

---

## 5. Client Engagement, Global Delivery & IP Safeguards

- **Headquarters**: Mumbai, Maharashtra, India.
- **Global Delivery Model**: Overlapping working hours for North America (EST/PST), Europe (GMT/CET), and APAC.
- **Turnaround Velocity**: Rapid 5-day client onboarding, 2-week MVP agentic sprint delivery, and bi-weekly milestone reviews.
- **Intellectual Property**: 100% full IP transfer on milestone delivery. Comprehensive bilateral Non-Disclosure Agreements (NDAs).
- **Communication Channels**: Dedicated Slack / Microsoft Teams shared channels, daily asynchronous standups, and weekly Zoom/Google Meet engineering synchronization.

---

## 6. Official Contact & Protocol

- **Website**: https://sicarius.in
- **General & Business Inquiries**: hello@sicarius.in
- **Technical Inquiries**: naman@sicarius.in
- **Direct Phone / Office Line**: +91 11 7122 8818
- **Security Vulnerability Reporting**: security@sicarius.in
- **Canonical Address**: Cyber City Tech Corridor, Mumbai, Maharashtra 400001, India