# Vahid Faraji, Applied AI / Researcher-Builder (Full Context)

> Applied AI Specialist shipping production agent systems and publishing at ACM. Multi-agent systems, context engineering, open source.

## Identity

- Name: Vahid Faraji
- Title: Senior Applied AI Specialist
- Organization: Kariyer.net (ilab group), Istanbul, Turkey
- Focus: Applied AI, Multi-Agent Systems, Context Engineering, MCP, LLM Systems
- Website: https://vfaraji89.github.io/
- Email: vfaraji89@gmail.com

## Professional Summary

Vahid Faraji is a Senior Applied AI Specialist with 9+ years of experience shipping production systems. He bridges research and production, publishing peer-reviewed papers at ACM while deploying enterprise-scale multi-agent systems. His work focuses on context engineering, MCP-based agent architectures, and open-source AI tooling.

## Publications (3 papers + 1 patent pending)

### ACM CAIS '26 (Peer-Reviewed)
- Title: Multi-Agent Position Classification with Tool Orchestration: Use Case System for Occupational Taxonomy Mapping
- Authors: Vahid Faraji et al.
- Venue: ACM Conference on AI and Agentic Systems, San Jose, May 2026
- DOI: 10.1145/3786335.3813207
- URL: https://dl.acm.org/doi/10.1145/3786335.3813207
- Page: https://vfaraji89.github.io/papers/cais-2026.html
- Key results: 95% accuracy on 14K+ positions, 72.5% faster classification, 41.8% fewer API calls
- Architecture: MCP-based multi-agent, planner, search, extraction, evaluation agents with shared caching

### arXiv: Tokalator
- Title: Tokalator: A Context Engineering Toolkit for AI Coding Assistants
- Author: Vahid Faraji
- ID: arXiv 2604.08290, April 2026
- URL: https://arxiv.org/abs/2604.08290
- Page: https://vfaraji89.github.io/papers/tokalator.html
- Contribution: Open-source toolkit for token-aware context management, Cobb-Douglas output-quality models, caching break-even analysis

### arXiv: Skill Extraction
- Title: Leveraging LLMs For Turkish Skill Extraction
- Authors: Vahid Faraji et al.
- ID: arXiv 2601.22885, January 2026
- URL: https://arxiv.org/abs/2601.22885
- Page: https://vfaraji89.github.io/papers/skill-extraction.html
- Contribution: First systematic LLM-based benchmark for Turkish NLP in workforce analytics

## Projects

### awesome-applied-ai: Applied AI Taxonomy
- URL: https://vfaraji89.github.io/awesome-applied-ai/
- GitHub: github.com/vfaraji89/awesome-applied-ai
- Type: Open source, curated reference
- What it is: A working map of applied AI for the enterprise. Six layers: Guardrails and Governance, Evaluation and Observability, Orchestration, Caching, Memory, and Retrieval. Includes a live tool index, equations, dictionary, commands, design problems and research findings.
- Scope: The tools, protocols, benchmarks and regulation that decide what a system knows, what it costs, and what you can prove about it.

### Tokalator: Context Engineering Toolkit
- URL: https://tokalator.wiki
- GitHub: github.com/vfaraji89/tokalator
- Type: Open source (MIT)
- Stack: Next.js, TypeScript, VS Code Extension, MCP, Token Economics
- Metrics: 500+ VS Code installs, 17 models supported, 58 wiki articles
- What it does: Real-time token budgeting, tab relevance scoring, Cobb-Douglas output-quality models, caching break-even analysis across 17 LLMs
- Programme: leads the Perplexity AI Business Fellowship

### Multi-Agent Job Classification System
- Type: Research + production
- Stack: MCP, LangGraph, Subagent Architecture, AI Search, Caching
- Metrics: 95% accuracy, 14K+ positions, 72.5% faster, 41.8% fewer API calls
- Published: ACM CAIS '26

### ProjectLens: Pricing Internal Tool (Kariyer.net)
- Type: Enterprise, internal
- Shipped: August 2026
- Team: Applied AI x Pricing & Revenue
- Stack: Applied AI, Microsoft Teams integration, Azure DevOps integration, SQL, workflow automation
- What it does: Takes a project request and manages the whole process end to end: effort calculation, pricing, proposal preparation, approval stage and follow-ups. Replaces a workflow where every pricing run meant a new Excel file.
- Notable: Read-only SQL console (only SELECT and WITH execute, each query runs in a rolled-back transaction on its own connection, nothing is writable) and a presentation mode that masks figures during screen shares.
- Metrics: 91% of manual effort eliminated; 64h manual baseline reduced to 5.5h; 59 hours freed (7.3 working days); 57h saved on pricing across 31 runs; 1.7h saved on request intake across 4 requests; 31 Excel files avoided.
- Measurement caveat: savings are modelled as 31 pricing runs x (2h - 10min) + 4 intake requests x (30min - 5min). The 2h manual baseline is a team estimate dated 2026-07-30, not an instrumented measurement.

### Enterprise AI Platform (Kariyer.net)
- Type: Enterprise, internal
- Stack: FastAPI, LangGraph, Agno, Pydantic, Arize
- Metrics: 60% workload reduction, 200+ users, group-wide AI standard
- What it does: Multi-agent orchestration for text-to-SQL, report automation, AI feedback loops

### 300M+ Datapoint Metadata Platform
- Type: Data infrastructure
- Stack: DataHub, Azure Synapse, BigQuery, Real-time Pipelines, Retool
- Metrics: 300M+ datapoints unified
- What it does: Metadata architecture, data governance, foundation for all AI/LLM deployments

### Baachee.ai: Data Agent Platform
- URL: https://baachee.ai
- Role: Technical Advisor (2026 - now), concurrent with the Kariyer.net role
- Stack: Agno (Self-Learning), Text-2-SQL, Metadata, Memory, Orchestrator
- What it does: Data agents, text-to-report pipelines, canvas flow tools, enterprise automation

## Experience

- 2025–now: Senior Applied AI Specialist, Kariyer.net (ilab group)
- 2026–now: Technical Advisor, Baachee.ai (advisory, concurrent)
- 2022–2025: Product Manager, Data, Kariyer.net
- 2021–2022: Business Analyst, WorqCompany
- 2019–2021: Co-Founder, Defaro.io (labor-market analytics startup)

## Memberships

- ACM Professional Member (Association for Computing Machinery)

## Certifications

- Google Generative AI Leader (Dec 2025, expires Dec 2028)
- Anthropic Model Context Protocol (Apr 2026)
- Anthropic Introduction to Agent Skills (Apr 2026)
- Anthropic Building with the Claude API (Apr 2026)
- Anthropic Claude Code in Action (Mar 2026)
- LangChain Introduction to LangGraph (Jan 2025)
- LangChain Introduction to LangSmith (Jan 2025)

## Teaching & Talks

- 18h Applied AI Course, Techcareer (2026): Custom tool design, contextual system architecture, multi-agent coordination
- AI Career Mentor, TEV Foundation (2026)
- Context Engineering for AI-Native Dev, Only Builders Club, Istanbul (2026)
- Deep Agent Frameworks for SaaS, Bridge Summit '25
- Enterprise Prompt Engineering, Kayseri University & Van University (2025)
- 91% manual effort removed and 59 hours freed (ProjectLens pricing tool)
- 400+ people trained, 50+ courses delivered

## Education

- M.A. Economics, Ibn Haldun University
- Thesis: Real-Time Labor Market Analysis
- Teaching Assistant: Econometrics I

## Languages

- Turkish: Native
- English: C1
- Spanish: Instructor level
- Portuguese: Basic
- Japanese: Learning

## Key Metrics

- 9+ years shipping production systems
- 95% agent accuracy (multi-agent classification)
- 72.5% faster classification
- 60% workload reduction (enterprise AI)
- 300M+ datapoints unified
- 3 publications + 1 patent pending
- 400+ people trained
- Perplexity AI Business Fellowship (programme lead)

## Hiring Summary

- Open to: senior applied AI roles (full-time), advisory engagements, enterprise AI training
- Role fit: AI Solutions Architect, Agentic AI Architect, Senior Applied AI Specialist, AI Product Manager, AI Platform Architect, AI Researcher, R&D Researcher
- Centre of gravity: solution and spec design for agent systems - framing the problem, architecting orchestration, owning the data and context layer
- Strengths: spec and solution design, agent architecture, context and token strategy, metadata and RAG readiness, ROI framing for business stakeholders
- Seniority: Senior / Principal, 9+ years across data products, solution design and applied AI
- Location: Istanbul, Turkey (GMT+3), remote-friendly with distributed teams
- Core stack: MCP, LangGraph, Agno, Vercel AI SDK, Next.js, TypeScript, Python/FastAPI, Anthropic SDK, Arize, BigQuery, Azure, DataHub
- Proof: ACM Professional Member, peer-reviewed at ACM CAIS '26, 2 arXiv papers, 1 patent pending, Perplexity AI Business Fellowship programme lead, Tokalator with 500+ installs
- Condensed hiring view: https://vfaraji89.github.io/?hire
- Contact: vfaraji89@gmail.com

## Social & Links

- GitHub: https://github.com/vfaraji89
- LinkedIn: https://linkedin.com/in/vahid-faraji-jobehdar
- X: https://x.com/vfaraji89
- Medium: https://medium.com/@vfaraji89
- Hugging Face: https://huggingface.co/vfaraji89
- Substack: https://substack.com/@vfaraji899
- Tokalator: https://tokalator.wiki
- awesome-applied-ai: https://vfaraji89.github.io/awesome-applied-ai/