# Pricing & Engagement Guide - Ibrahim Salman (Freelance Full-Stack & AI Engineer)

## Pricing Philosophy

Scope-based pricing, agreed before work starts. Because RAG architectures, OCR pipelines, and automation backends vary significantly in complexity and data scale, projects are quoted based on defined scope and clear milestones rather than arbitrary hourly billing.

## Engagement Tiers

1. **Fixed-Scope Deliverables**
   - Ideal for: Turnkey AI assistants (RAG), custom OCR/document intelligence pipelines, marketplace scrapers with alerting, or full-stack Next.js web applications.
   - Structure: 50% on kickoff, 50% upon verified deployment and handover.
   - Deliverables: Production-ready codebase, complete automated test suite (Pytest/Vitest/Playwright), infrastructure deployment (Vercel/Cloud), and documentation.

2. **Sprint-Based / Ongoing Retainer**
   - Ideal for: Fast-moving teams needing dedicated engineering capacity for ongoing feature development, pipeline scaling, model fine-tuning, or system maintenance.
   - Structure: Billed bi-weekly or monthly per sprint with agreed deliverables.

3. **Architecture Audits & Technical Consulting**
   - Ideal for: Teams needing code reviews, failure mode analysis, RAG retrieval optimization, OCR benchmark audits, or anti-scraping evaluations.

## What Determines the Quote
- **Scope & Functional Requirements** - Defined features, integrations, and success criteria.
- **Timeline & Urgency** - Standard sprint velocity vs. accelerated critical deadlines.
- **Data & System Complexity** - Single-endpoint utility vs. multi-source concurrent pipeline with auth, vector embeddings, multi-LLM fallbacks, and rate limiting.

## How to Get a Quote
Email **ibrahim.pk848@gmail.com**, submit the scoping form at **https://ibrahimsalman.vercel.app/hire**, or use **https://ibrahimsalman.vercel.app/contact**.
- **Location**: Islamabad, Pakistan (UTC+5 / PKT), working remotely.

## Shipped Work
- **UET GPT** - Unofficial RAG assistant for UET Taxila students (not affiliated with the university), with model fallbacks, live at https://uetgpt.site
- **B.L.A.S.T.** - OCR engine, public demo, 914 collected tests, character error rate 0.499 → 0.108 on a 14-page scanned-book test set
- **Marketplace Monitoring & Alerting** - Fail-closed monitoring system built during an internship at Paramount Intelligence, 304 collected tests across 4 Python versions × 2 OSes in CI

Full portfolio: https://ibrahimsalman.vercel.app/projects

## Current Availability
Available for new freelance projects, contract engagements, and consulting roles.

---
Last updated: 2026-10-24