Syed Zain Gillani // AI Engineer
--:--:-- GMT+5
AJK, PK → UAE // OPEN
0.00, 0.00
Portfolio / 2026 / Index

Syed ZainGillani

I build RAG systems, AI agents, and secure, low-latency LLM applications. Two years shipping the web, two years shipping intelligence on top of it.

Scroll
01

Who's writing the prompts

A self-taught engineer who went from markup to models, and now ships production AI for real businesses while finishing a CS degree.

I started with HTML, CSS and JavaScript in 2022, building sites for clients as a freelancer and at CodingEagles. When large language models got good, I followed the interesting problem: how do you make them accurate, fast, and safe enough to trust in a live product?

Since 2024 that has been the whole job. I design retrieval pipelines that actually retrieve, agents that do work instead of hallucinating it, and deployments hardened against prompt injection and built to answer in milliseconds, not seconds. I work across Claude and Google AI SDKs, fine-tune open models when a task needs it, and wire everything together with MCP, n8n and a Linux box I am not afraid of.

Based in Azad Kashmir, Pakistan. Open to relocating to the UAE, available immediately.

Syed Zain Gillani
SZG / 2026Muzaffarabad
02

What I reach for

Generative AI

// core
  • RAG pipelines
  • Context engineering
  • Prompt engineering
  • Embeddings
  • Hybrid retrieval
  • Reranking
  • Fine-tuning · LoRA
  • Evals · RAGAS

Agents & Orchestration

// autonomy
  • Multi-agent systems
  • Function calling
  • MCP
  • Structured outputs
  • LangChain
  • LangGraph
  • Tool routing

Models & SDKs

// providers
  • Anthropic Claude
  • Google AI · Gemini
  • Open-weight LLMs
  • Custom deployment
  • Self-hosted serving
  • Latency tuning

Systems & Web

// foundation
  • Python · FastAPI
  • HTML · CSS · JS
  • REST APIs
  • Docker
  • Linux servers
  • PostgreSQL · pgvector

Automation

// glue
  • n8n
  • Zapier
  • Workflow design
  • Webhooks
  • CRM & messaging integration

Security & Ops

// trust
  • Prompt-injection defense
  • Input validation
  • Rate limiting
  • Secrets hygiene
  • Monitoring
03

Selected work

W-01

CodingEagles internal AI platform

The system the agency runs on. One interface unifying document search, task agents and workflow automation, used across day-to-day operations.

Claude SDKGoogle AIFastAPIn8nDocker · Linux
2024 → Present
◆ Deployed
W-02

Infrastructure RAG pipeline

Retrieval-augmented assistant over internal docs and infrastructure. Chunking, embeddings, hybrid retrieval and reranking so answers are sourced, not guessed. Lookups that took minutes now take seconds.

Embeddingspgvector · QdrantRerankingLangChain
Internal
◆ RAG
W-03

Restaurant ordering agent

Conversational agent for local restaurants: menu Q&A, order taking and customer support, wired to live business data through function calling and MCP. Answers on WhatsApp before a human would have picked up.

Function callingMCPLow-latency serving
Local business
◆ Live
W-04

Clinic booking assistant

An appointment assistant for a local medical clinic. It reads availability, books and reschedules over chat, and hands off to staff the moment a question needs a human. Front desk stopped drowning in the same five questions.

Agent + toolsCalendar APIGuardrails
Local business
◆ Live
W-05

Hivly web tools

A suite of fast, no-nonsense web tools I designed and built, all living under one roof at hivly.net. Shipped and maintained end to end, front end to deploy.

JavaScriptHTML · CSSResponsive UIProduct
Web · Product
↗ hivly.net
W-06

Enterprise document intelligence

A document-intelligence system for a client in the legal-and-compliance space, extracting and answering across a large private corpus. Specifics are covered by an NDA.

RAGPrivate deployment████████████
Under NDA
◆ Confidential
CONFIDENTIAL
W-07

Fraud-signal LLM pipeline

A detection and triage pipeline for a fintech operating in the region, combining embeddings with an LLM judge over transaction streams. Client, metrics and architecture withheld under NDA.

EmbeddingsLLM-as-judge██████████
Under NDA
◆ Confidential
CONFIDENTIAL
04

How I ship

P/01

Secure by default

Prompt-injection defenses, input validation and secrets hygiene are part of the first commit, not a later audit.

P/02

Fast where it counts

Latency is a feature. I tune retrieval, caching and serving so a model answers in the time a user will actually wait.

P/03

Shipped, then measured

A demo is not a product. I deploy real systems, watch them, and let evals decide what changes next.

05

Let's build something

zain@codingeagles.dev ↗
Phone
+92 325 1736468
CV
Download PDF
Based
Muzaffarabad, AJK, Pakistan
Languages
EN · UR · PA · Hindko
Open to relocation · UAE
© 2026 Syed Zain Gillani Built by hand · No templates zain.codingeagles.dev