Skip to content
AboutAI LabWorkWritingStackContactHire me
Available for freelance & contract work

Atif Iqbal — Full-Stack & AI Engineer in Lahore, Pakistan

>

I build production web platforms and the AI systems that make them think — RAG pipelines, vector search, small language models and autonomous agents.

View my workExplore the AI Lab
atif4aws@gmail.com+92 333 8340143Lahore, Pakistan
Atif Iqbal, Full-Stack & AI Engineer
RAG
Vector DB
Next.js
Agents
10+years shipping
production software
Next.jsReactTypeScriptLaravelPHPNode.jsPythonRAGEmbeddingspgvectorQdrantPineconeLangChainMCPSLM / LoRAVector searchBladeStripeHubSpotAWSDockerMySQLRedisWeb3Next.jsReactTypeScriptLaravelPHPNode.jsPythonRAGEmbeddingspgvectorQdrantPineconeLangChainMCPSLM / LoRAVector searchBladeStripeHubSpotAWSDockerMySQLRedisWeb3
About me

Engineering the web, teaching it to think

A decade of shipping client platforms, now paired with the AI systems that make them smarter.

Atif Iqbal — Full-Stack & AI Engineer
Atif Iqbal
Full-Stack & AI Engineer
Lahore, Pakistan

I have spent the past decade building and shipping production software for clients across the UK, the USA, Australia, the UAE and Pakistan — booking marketplaces, learning platforms, cryptocurrency exchanges, payment products and multi-tenant SaaS. My work sits at the architecture level: the data model, the API surface, the infrastructure it runs on — then owning the build through launch and everything that comes after it.

For the past few years that practice has extended into applied AI. I design retrieval-augmented generation pipelines over private corpora, tune embedding and chunking strategies against real query sets, run vector databases under production load, fine-tune small language models where cost and latency matter more than raw capability, and build agentic commerce systems where an agent — not a form — completes the journey.

What has not changed is the standard. Systems that hold up under real traffic, integrate cleanly with how the business actually operates, and are handed over documented and maintainable. I care far less about which framework is fashionable this year than about whether the thing is still earning its keep two years after launch.

Architecture & delivery

System design, data models and API contracts — then owning the build from first commit through launch and scale.

Applied AI

RAG pipelines, vector search, fine-tuned small models and agents that complete real tasks, not demos.

Commercial judgement

Payment rails, CRMs and multi-tenant SaaS designed around the business outcome, not the tech stack.

0
Projects shipped
0
AI systems built
0
Industry domains
0+
Years engineering
AI Lab

Retrieval, agents and models in production

The AI half of my practice: grounded retrieval over private data, vector search that holds up under load, small models tuned for cost, and agents that finish the job.

Anatomy of a RAG pipeline

The path a question takes from your documents to a grounded answer.

live flow
chunk overlaptop-krecall@kcross-encodercitationsrefusal policy
Retrieval-Augmented Generation

RAG Systems

Assistants that answer from your data, not from memory. I build the whole path — ingestion, chunking, retrieval, reranking and grounding — so answers arrive with citations and the model says nothing it cannot source.

  • Hybrid search: BM25 keyword + dense vector recall
  • Semantic and recursive chunking with overlap tuning
  • Cross-encoder reranking for precision at the top
  • Citation spans and answer-grounding checks
  • Query rewriting, HyDE and multi-query expansion
LangChainLlamaIndexHybrid searchReranking
Semantic representation

Embeddings

Retrieval quality is decided long before the LLM is called. I choose and evaluate embedding models, tune chunk geometry, and measure recall on real queries rather than guessing.

  • Model selection across dimensions and cost tiers
  • Multilingual and domain-adapted embeddings
  • Chunk size, overlap and metadata design
  • Recall@k evaluation against golden query sets
  • Semantic caching and deduplication
OpenAICohereBGEE5Matryoshka
Search infrastructure

Vector Databases

Running vector search in production means index tuning, metadata filtering and honest latency budgets. I ship it on Postgres when that is enough, and on a dedicated engine when it is not.

  • pgvector, Pinecone, Qdrant, Weaviate and Chroma
  • HNSW and IVF-Flat index tuning for recall vs latency
  • Metadata filtering and multi-tenant isolation
  • Sharding, replication and re-index strategies
  • Cost modelling per million vectors
pgvectorQdrantPineconeHNSW
SLM fine-tuning & serving

Small Language Models

Not every task needs a frontier model. Fine-tuned 3B–7B models handle narrow, high-volume work at a fraction of the cost — and they can run on your own hardware.

  • LoRA / QLoRA fine-tuning on curated datasets
  • 4-bit and 8-bit quantization (GGUF, AWQ)
  • vLLM serving with continuous batching
  • On-prem and edge deployment for data residency
  • Confidence-based routing to larger models
QLoRAvLLMGGUFDistillation
Agents that transact

Agentic Commerce

Commerce is moving from pages to agents. I build the tool layer — catalogue, cart, checkout — that lets an agent complete a purchase safely, with a human in the loop where it counts.

  • MCP servers exposing storefront capabilities as tools
  • Tool-calling agents for discovery, cart and checkout
  • Agent-readable catalogues and structured product feeds
  • Guarded payment steps with explicit confirmation
  • Headless storefront and payment-gateway integrations
MCPTool callingCatalogueCheckout
Making it trustworthy

Agents, Evals & Guardrails

Demos are easy; reliability is the work. Multi-agent graphs with typed handoffs, golden-set evaluation on every change, and guardrails against injection, leakage and quiet regressions.

  • Planner / executor and multi-agent graphs
  • Golden datasets and LLM-as-judge scoring
  • Prompt-injection and PII guardrails
  • Tracing, replay and token-cost observability
  • Human-in-the-loop approval gates
LangGraphEvalsTracingGuardrails
Selected work

Platforms shipped, problems solved

23 builds across AI, SaaS, fintech, commerce, education and healthcare — from booking marketplaces and mining ERPs to multi-agent clinical retrieval and on-chain checkout.

Writing

Notes from the build

What I have learned shipping Laravel platforms and AI systems — retrieval, agents, small models and the unglamorous parts that decide whether any of it works.

All 8 articles
Capabilities

The stack behind the work

Tools I reach for daily — across the classic web stack, the AI layer and the infrastructure that keeps both running.

AI Engineering

RAG pipelinesEmbeddingsVector searchSLM fine-tuningLoRA / QLoRAAgent orchestrationTool calling / MCPPrompt engineeringEvals & guardrails

Vector & Data

pgvectorPineconeQdrantWeaviateChromaElasticsearchMySQLPostgreSQLRedis

Frontend

TypeScriptReact.jsNext.jsTailwind CSSReduxComponent architectureResponsive UIWeb performance

Backend

PHPLaravelCodeIgniterNode.jsPythonREST APIsWebSocketsQueues & workersAuth & RBAC

Commerce & Integrations

StripePayment gatewaysSubscriptions & billingHubSpot CRMThird-party APIsWebhooksHeadless commerceSEO

Cloud & DevOps

AWSVPSLinux (CentOS)DockerNginxApacheGitCI/CDMonitoring

Web3 & Crypto

Web3.js / ethersWallet integrationExchange APIsTrading enginesMarket data streamsSmart contract integration

Domains I have built in

01

RAG Knowledge Systems

Grounded assistants over private corpora with citations and guardrails.

02

Agentic Commerce

Agents that search catalogues, build carts and complete checkout.

03

E-commerce Platforms

Catalogues, checkout, inventory sync and payment gateways at scale.

04

Point-of-Sale Systems

Fast, offline-tolerant retail terminals with real-time sync.

05

Learning Management (LMS)

Course delivery, certifications, student portals and progress tracking.

06

Booking & Management

Scheduling, availability, payouts and automated reminders.

07

ERP Systems

Inventory, procurement and reporting for operations teams.

08

Trading Bots

Signal engines, execution loops and exchange integrations.

09

Cryptocurrency Exchanges

Order books, wallets, KYC flows and market data streams.

10

Smart Contracts

On-chain logic, wallet flows and Web3 integrations for token products.

11

SaaS Systems

Multi-tenant products with billing, roles and analytics.

Get in touch

Have something worth building?

Whether it is a web platform, a RAG assistant over your own data, or an agent that can actually transact — tell me what you have in mind.

Based in
Lahore, Pakistan
Working with clients worldwide

How we would work together

Taking work now
  1. 01

    Discovery

    A short call to understand the problem, the constraints you are working under and what success actually looks like.

  2. 02

    Scope & proposal

    A written scope with milestones, a timeline and a fixed price or day rate — so there are no surprises later.

  3. 03

    Build in sprints

    Weekly demos against a live staging environment. You see working software, not status reports.

  4. 04

    Handover & support

    Documented code, a deployment runbook and a support window after launch. The system stays yours.

Helps me reply faster

What you are buildingRough budgetTimelineExisting stack
Email meWhatsApp