Websites that bring in business
A fast company website you can update yourself. I set up Search Console, a sitemap, a blog and analytics with cookie consent, so people can find you on Google.
- Next.js
- SEO
- GA4
Waqar Ahmed: From the database to the domain.
01About
About Waqar Ahmed
I turn the way your business already works into software your team uses every day, and I keep it running long after launch. If AI can take repetitive work off their hands, I build that in too.
waqar@islamabad:~ $ whoami --long
waqar@islamabad
- role
- Full-stack developer
- builds
- Web apps, internal tools, AI features
- handles
- Database · servers · security
- location
- Islamabad, PK (UTC+5)
- uptime
- Building since 2021
- status
- Open to work
- FastAPI
- Next.js
- PostgreSQL
- React
- Docker
- Cloudflare
- TypeScript
- Cloud Run
- AWS
- Railway
- GitHub Actions
- Vite
02Services
What I can build for you.
You work with one person for the whole project, hosting included. These are the things I build most often.
01Website01 / 05 02Business software02 / 05 Software your team uses every day
CRMs, billing, inventory, quotations and ledgers, built around the way your team already works and checked against your real invoices. Each person only sees what their role allows.
- FastAPI
- React
- PostgreSQL
03Integrations03 / 05 Integrations and automation
Leads from Meta ads and your website go straight into your CRM. I also set up live notifications, Stripe payments and yearly archives to your Google Drive.
- Webhooks
- Stripe
- Google Drive
SeeCRM 2.0
04Hosting & security04 / 05 Hosting and security
Your app runs in Docker on Cloud Run, AWS or Railway, behind Cloudflare with a firewall and rate limits. If you already have an app, I can audit its security and tell you what to fix first.
- Docker
- Cloudflare
- AWS
05AI features05 / 05 AI inside your app
An assistant that answers questions from your own documents, an agent that handles repetitive writing, or a fine-tuned model. It runs as part of your app and gets deployed with it.
- LangGraph
- RAG
- Fine-tuning
SeeRAG Bee
03Selected work
Shipped and running.
Most of these are used every day by the businesses they were built for, like the CRM, the real estate platform and the billing system. There are a few AI builds in here too.
Family Realtors PK
Listings, development projects and a blog for a real estate company, run from its own admin dashboard.
Pages are statically generated and revalidated the moment an admin or writer edits them. Visitors get instant search with filters, project galleries with floor plans, and Google sign-in to save favourites or submit a listing for approval. The hero is a scroll-driven three.js key traced from the logo. Next.js 16 and FastAPI on PostgreSQL.
- Next.js 16
- React 19
- three.js
- FastAPI
- PostgreSQL
- Railway
CRM 2.0
A real estate CRM where every company gets its own database and an AI assistant that sees only what its user can.
Approving a signup creates the company's own PostgreSQL database and migrates it. Three roles (super admin, manager and agent) scope every query, including those of Echo, a read-only LangGraph assistant. Leads arrive from Meta Lead Ads by webhook and trigger WhatsApp alerts, and notifications reach the browser live over SSE. FastAPI and React / TypeScript, with files on Cloudflare R2.
- FastAPI
- React
- TypeScript
- LangGraph
- PostgreSQL
- Meta Lead Ads
Billing & Inventory System
Private software that runs the invoicing, stock and ledgers of a wholesale distributor, verified to the rupee.
Quotations, invoices, dispatch notes and returns are typed into a spreadsheet-style grid with a quantity per colour, and each one posts to category-scoped dealer and vendor ledgers. Three-stage discounts and moving-average costing run across 1,000+ products and 3,500+ variants.
- FastAPI
- PostgreSQL
- React
- TypeScript
- PWA
RAG Bee
Answers grounded in a library of books, routed by an agent and served from my own GPUs.
LangGraph routes each query. Self-hosted on Modal, serving Qwen 8B on L4 GPUs with BGE-Large embeddings and a BGE reranker for hybrid search. Containerised for Railway and Koyeb.
- LangGraph
- Qwen 8B
- BGE
- Modal
- FastAPI
- React
The Magic Lamp
A webapp for a technology, media and IP company in Kuwait, built around a lamp you can swing.
A three.js filament lamp hung on a damped pendulum: drag it to swing it, tap the bulb for sparks. GSAP and Lenis drive the scroll, with split-text headlines, a cord threaded through the process steps, a phone cycling through the flagship app, and a dotted world map that marks Kuwait and the GCC. Without WebGL the lamp falls back to an SVG drawing.
- Next.js 16
- React 19
- three.js
- GSAP
- PostgreSQL
Zeeshan Saleem
A portfolio site for a real estate advisor and builder in Rawalpindi and Islamabad.
A WebGL city at dusk in the hero, six houses that rise in 3D stage by stage as you scroll, and a contact form that feeds a private enquiry inbox.
- Next.js
- three.js
- PostgreSQL
LinkedIn Scraper API
Profile data as structured JSON, in real time, over plain HTTP.
Keeps persistent Chrome sessions behind a one-time authentication, extracts LinkedIn profile data on request and returns clean, structured JSON.
- Python
- REST
- Chrome
- Selenium
04Process
How we’d work together.
The same four stages whether it’s a company website or a full CRM. You see working software early, and nothing goes live until it’s checked.
Stage 01
Talk
You tell me how the business runs today and what keeps getting in the way. I look at what you already have before suggesting anything new.
Stage 02
Plan
I write up a short plan covering what gets built first, where it will run and what I need from you. We start with a small first version and build from there.
Stage 03
Build
You get a live link early on, so you can click through it yourself. I test it against your actual data, like invoices, leads and stock.
Stage 04
Launch
I deploy it behind Cloudflare and set up backups. After launch we can keep going with fixes and new features.
05Stack
Layer by layer.
The tools I use day to day.
L0 Edge & security
- Cloudflare
- DNS
- WAF
- Rate limiting
- Turnstile
L1 Frontend
- React
- Next.js
- TypeScript
- Vite
- PWA
L2 Backend
- FastAPI
- Python
- Django
- Alembic
- pytest
L3 Data
- SQL
- PostgreSQL
- Cloudflare R2
L4 Infrastructure
- Docker
- GitHub Actions
- GCP Cloud Run
- AWS ECS
- Railway
I also know CCNA-level networking.
06Experience
How I got here.
I started out training ML models for clients on Upwork. Most of the work that came after was web development and deployment, and that’s my main work now.
Role 1 of 4
Machine Learning Engineer
- org
- Upwork
- where
- Remote
- when
- 2021 - 2023
Client work included a CNN chess evaluator with minimax and alpha-beta pruning, deep learning to detect and categorise ads on live TV, and DBSCAN-based sports odds analysis that flags bookmaker discrepancies.
Role 2 of 4
Freelance Developer
- org
- Self-employed
- where
- Remote
- when
- Mar 2023 - Feb 2026
Built and shipped web apps for clients in Pakistan and abroad, usually as the only developer. Mostly FastAPI or Django on the backend, React on the frontend, PostgreSQL and Docker.
Some projects needed AI for content generation, real estate or data extraction, so I built LangGraph agents, RAG pipelines and fine-tuned LLMs for them.
Role 3 of 4
Private Tutor
- org
- One-on-one, in person
- where
- Rawalpindi
- when
- 2024 - Feb 2026
Taught CCNA networking, data science fundamentals, full-stack web development and advanced mathematics (calculus, linear algebra) with customised learning plans.
Now · Part-time
Technical Team Lead
- org
- Family Realtors PK
- where
- Islamabad · Contract
- when
- Feb 2026 - Present
Hired on contract to build the company's website (familyrealtorspk.com) and a multi-tenant CRM with role-based access, live notifications and Meta Lead Ads integration.
Handled deployment and security end to end. The apps run on Cloud Run and Railway behind Cloudflare, with files on R2, and I ran a security audit with a tiered plan of fixes.
Now I work part-time, keeping both systems running and maintained.
Education & certification
- Microsoft Azure Data Scientist AssociateMicrosoft · January 2025Verify
- BS PhysicsCOMSATS University Islamabad
07Lab
Where the AI comes from.
Machine learning is where I started, and I still train and fine-tune models. These are client projects, research and open-source work from that side of things.
Fig. 1
Two chess engines
A CNN evaluator searched with minimax and alpha-beta pruning, built for a client. And an open-source ensemble of an encoder-only transformer, UCB exploration and Monte Carlo tree search that plays in the terminal.
Fig. 2
Ads on live TV
Deep learning that detects where commercial breaks start in a live broadcast and sorts each ad into a category. The pipeline is documented; the source is under NDA.
Fig. 3
Reading the odds
DBSCAN clusters bookmaker prices to find the consensus, then flags a bookmaker as mispriced when its variance from the cluster passes a threshold.
Fig. 4
BERT, fine-tuned
BERT fine-tuned twice: once for sentiment analysis, once to grade essays.
08Contact
Need something built?
I’m open to new projects right now. Tell me what you need with the form and it comes straight to me. If you’d rather talk, email or WhatsApp work just as well.
Islamabad PKT · UTC+5