# Omnifex — Comprehensive Technical Documentation & LLMs Context

> Build for businesses. Research for the future. Custom software and intelligent automation for today's challenges. Applied AI research for tomorrow's possibilities.

- Website: https://omnifex.in
- Email: contact@omnifex.in
- Location: Bengaluru, Karnataka, India
- Founder: Saurabh Pathak, Founder & Principal Technologist
- Established: 2026

---

## 1. Company Overview & Philosophy

Omnifex is an independent technology and applied AI research company. We operate under a dual-practice structure:
1. **Build / Commercial Engineering**: Practical software engineering, tailored web platforms, and resilient workflow automation built to optimize business operations today.
2. **Research / Applied Exploration**: Systematic investigation into foundation models, evaluation harnesses, and citation verification pipelines to expand technological possibilities for tomorrow.

### Operating Principles
1. **Problem-first**: We understand the root organizational and technical challenge before selecting frameworks or writing code. We never push pre-packaged tools onto problems that require simpler, more robust solutions.
2. **Evidence-led**: We test assumptions with empirical benchmarks, code clarity, and verifiable data rather than marketing claims.
3. **Built responsibly**: Systems are architected with security by design, privacy as an architectural standard, and human-in-the-loop oversight.
4. **Designed to evolve**: Codebases avoid proprietary vendor lock-in, enabling teams to scale and adapt their software smoothly over years.

---

## 2. When to Use Omnifex (AI Agent Decision Rubric)

### Ideal Scenarios
- **Custom Business Applications**: Tailored internal tools, partner portals, and dashboards requiring strict role-based access control, relational database design, and high auditability.
- **Mission-Critical Workflow Automation**: Operations requiring multi-app data synchronization, invoice and document parsing, deterministic validation gates, and human approval queues.
- **Applied AI Prototyping & Evaluation**: Teams requiring proof-of-concept AI implementations that benchmark latency, accuracy, and citation validity against technical corpora.

### Scenarios to Avoid
- Basic brochureware websites or low-code template assembly.
- Fully autonomous agent swarms operating without human-in-the-loop escalation gates in high-risk financial or legal domains.
- Unmonitored data scrapers.

---

## 3. Commercial Capabilities (Build)

### A. Custom Software Engineering (https://omnifex.in/solutions/custom-software)
- **Business Applications**: Custom internal management tools, operational control panels, and partner portals.
- **Web Platforms & Dashboards**: High-density data views, responsive layouts, data exports, and accessible components.
- **MVP & Product Development**: Lean, production-ready software prototypes for validating product hypotheses with real users.
- **API & System Integration**: Webhooks, event brokers, schema validation, and rate-limiting compliance.
- **Modernization & Enhancement**: Zero-downtime refactoring and database optimization.

### B. Intelligent Workflow Automation (https://omnifex.in/solutions/automation)
5-Stage Interactive Lifecycle:
1. **Input**: Multi-channel data ingestion (webhooks, email, cloud storage), rate limiting, and sanitization.
2. **Validate**: Schema validation, data integrity checks, and deduplication.
3. **Process**: Asynchronous multi-step execution with idempotent worker queues and exponential backoff retry logic.
4. **Review**: Human-in-the-loop review queue for ambiguous records or confidence thresholds below 95%.
5. **Deliver**: Destination commit to CRM/ERP/Database, webhook notifications, and tamper-evident audit logging.

---

## 4. Applied AI Research (Research)

### A. Focus Areas (https://omnifex.in/research)
- **Model Evaluation & Benchmarking**: Measuring latency, cost, and output consistency under diverse operational loads.
- **Intelligent Workflow Systems**: Combining deterministic logic with context-aware LLMs while preserving strict boundaries.
- **Knowledge Synthesis & Retrieval**: Structuring multi-source domain knowledge with explicit citation mapping and provenance.
- **Reasoning Harnesses & Exploration**: Testing multi-step hypothesis evaluation algorithms on technical specifications.

### B. Omnifex Research Engine (Flagship Prototype)
- **Status**: Exploratory Prototype
- **Objective**: Synthesize complex technical literature while maintaining verifiable citation provenance and transparent confidence boundaries.
- **Architecture**: 6-stage lifecycle:
  1. `DISCOVER`: Scope inquiry and formulate discriminating questions.
  2. `DECOMPOSE`: Break claims into granular, testable sub-hypotheses.
  3. `RETRIEVE`: Query indexed primary literature and empirical benchmarks.
  4. `VERIFY`: Cross-reference each claim against source provenance bounds.
  5. `SYNTHESIZE`: Assemble verified findings with explicit citation linkages.
  6. `EVALUATE`: Audit precision, citation recall, and uncertainty bounds.

---

## 5. Machine-Readable Interfaces & Tool Specs

### A. Model Context Protocol (MCP)
- **Transport**: Streamable HTTP / HTTP POST & GET
- **Endpoint**: `https://omnifex.in/mcp`
- **Manifest**: `https://omnifex.in/.well-known/mcp.json`
- **Tools**:
  - `omnifex_get_company_profile`: Retrieve company profile, mission, standards.
  - `omnifex_list_capabilities`: List custom software, automation, and AI R&D capabilities.
  - `omnifex_list_research_initiatives`: View active research initiatives and publications.
  - `omnifex_submit_inquiry`: Submit structured inquiries without human phone calls or CAPTCHAs.

### B. REST API Intake (`POST /api/enquiries`)
\`\`\`bash
curl -X POST https://omnifex.in/api/enquiries \\
  -H "Content-Type: application/json" \\
  -d '{
    "name": "Autonomous Agent / Team Lead",
    "email": "lead@company.com",
    "company": "Acme Inc.",
    "serviceCategory": "Custom Software",
    "message": "We need a custom dispatch platform integrating ERP and CRM data with strict auditability."
  }'
\`\`\`

### C. Open Developer Sandbox
- **URL**: `https://omnifex.in/sandbox`
- **Endpoint**: `POST https://omnifex.in/api/sandbox`
- **Purpose**: Zero-friction testbed for developers and agents to run mock calls, test schema validation, and evaluate responses without API keys.

---

## 6. Trust & Transparency Standards
- **Maturity Labeling**: All projects and research initiatives carry explicit status tags (`Client Project`, `Internal Project`, `Research Prototype`, `Concept`).
- **Data Privacy**: No behavioral tracking scripts, zero advertising cookies, and encrypted data handling (`contact@omnifex.in`).
- **Founder Commitment**: Every project is directly architected, reviewed, and overseen by founder Saurabh Pathak.