Zain Ul Abideen

⚡ AI Engineer & Software Engineer

Building Production LLMs, Multi-Tenant SaaS & Data Architectures

# Production RAG & Vector Intelligence Pipeline
async def query_intelligence(query: str, tenant_id: str) -> Response:
    embeddings = await embed_model.aembed(query)
    context = await qdrant.search(
        collection="tenant_knowledge",
        query_vector=embeddings,
        query_filter={"tenant_id": tenant_id},
        limit=5
    )
    return await llm.agenerate(query=query, context=context)

• 🇵🇰 Based in Lahore, Pakistan
• 🧠 AI Engineer & Software Engineer with 3+ years designing, building, and deploying production systems
• ⚡ Deep expertise in LLM applications, RAG pipelines, Document AI, Text-to-SQL, and vector retrieval
• 🏗️ Proven track record building multi-tenant SaaS platforms, automated financial engines & real-time analytics
• 📈 Shipped systems delivering $30M+ transaction pipelines, 60% operational overhead cuts, and 333% growth in acquisitions
• 📨 Reach me directly at [email protected]
• 🟢 Open to work — ready to build scalable, intelligent products!

Conversational BI SaaS turning natural-language questions into validated SQL, live charts, and interactive dashboards without requiring manual SQL queries.

• AST Security & Mutation Blocking: Model-generated SQL parsed via sqlglot AST; mutation nodes strictly rejected inside read-only transactions to protect production databases.
• Secure Remote Tunneling: SSH bastion with asymmetric trust & public-host SSRF defenses to securely query private-subnet databases.
• Cost Optimization: Compact schema serialization reducing LLM planning-stage token payload by ~60%.
• Stack: FastAPI • React 19 • TypeScript • SQLAlchemy 2.0 • PostgreSQL 16 • sqlglot • OpenAI / Gemini / Groq • Docker

Multi-tenant FinTech platform managing the entire loan lifecycle from inbound lead capture to funded deal, with AI embedded at every stage.

• AI Lead Scoring: Gemini-powered classifier scoring leads (0–100) with explainable quality and spam indicators.
• 2-Phase Document OCR: AWS Textract + Gemini pipeline for automated document classification, structured data extraction, and audit findings.
• AI Deal Master Summary: Synthesizes multi-document financials into an underwriting credit memo with risk and collateral analysis.
• Impact: 60% reduction in operational overhead, 30% faster loan processing, 25% revenue growth, and 100% visibility into case status.
• Stack: Django REST Framework • Google Gemini • AWS Textract • PostgreSQL • Celery + Redis • DocuSeal • Docker

Private lending automation platform replacing manual spreadsheet tracking with a comprehensive automated financial calculation and payment engine.

• Automated Fund Splitting: Stripe Connect handles multi-party payouts (borrower → seller → referral commission → platform fees) in a single flow.
• Amortization Engine: Exact principal and interest calculations per installment with real-time running ledger updates.
• Multi-Role Portals: Scoped dashboards for Sellers, Borrowers, Referrals, and Admins.
• Stack: Next.js • Django REST Framework • PostgreSQL • Stripe Connect • Celery • AWS EC2

End-to-end industrial platform automating iron rebar design, quotations, and confirmed ordering for the construction industry.

• Interactive 2D Drafting Engine: Built with Konva.js for parametric sketching of bars, nets, and cages with real-time measurement feedback.
• Automated Production Tickets: Machine vendors receive automated, production-ready tickets directly from order submissions.
• Recognition & Scale: Generated $30M+ in potential orders in year 1, 1,000+ active users in month 1, and selected for the Contech Incubator.
• Stack: React.js • Django • Konva.js • PostgreSQL • AWS Lambda + EC2 • Docker

Automotive intelligence platform leveraging Open Car Data (OCD), automated scraping, and market analytics for dealership inventory acquisition.

• Automated Data Harvesting: AWS Lambda + Selenium workers for continuous data harvesting, sync, and automated outreach via SendGrid.
• Sales Prospects Module: Algorithms analyzing vehicle market patterns to surface high-margin inventory acquisition leads.
• Impact: 333% surge in private-party acquisitions, 47% increase in customer traffic, 12% higher gross sales.
• Stack: Next.js • Django • Selenium • PostgreSQL • AWS Lambda + RDS + S3 • SendGrid

Modular Python pipeline utilizing OpenAI Whisper for accurate speech-to-text audio processing.

• Capabilities: High-fidelity multi-lingual transcription, support for various audio formats, and clean CLI execution.
• Stack: Python 3.11 • OpenAI Whisper • PyTorch

┌─────────────────────────────────────────────────────────────────┐
│                                                                 │
│   " Make it work, make it right, make it fast.                  │
│     In that order. Always. "                                    │
│                                               — Kent Beck       │
│                                                                 │
│   → Ship early, iterate often                                   │
│   → Vector search without good chunking is just noisy math      │
│   → An untested API is a broken API you haven't found yet       │
│   → If the deploy process scares you, automate it               │
│   → Good architecture > clever code                             │
│   → Logs, metrics, and evaluations are not optional             │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘