FAQ

Questions clients ask

How I work, what I build with, who owns the code, how engagements start, and which regions I serve. Answers for companies considering hiring me for AI or full-stack work.

About me

Who is Muhammad Sallar Bin Aamir?
Muhammad Sallar Bin Aamir is a Senior Full Stack AI Engineer and AI Researcher based in Islamabad, Pakistan. He builds production-grade full-stack and AI-powered systems for international clients across the US, Canada, the UK, the UAE, Brazil, and Australia, and has published peer-reviewed research in IEEE Access (2025).
What does Sallar specialize in?
Full-stack web development with Next.js, React, Node.js and FastAPI; AI engineering with LangChain, OpenAI, retrieval-augmented generation and agentic systems; and computer vision with PyTorch, OpenCV and YOLOv8. He has shipped products in QA automation, education, healthcare imaging, and mixed reality.
Where did Sallar study and what are his academic honors?
He earned a Bachelor of Science in Computer Science from Bahria University, Islamabad (2021–2025), graduating Magna Cum Laude with the Gold Medal and a CGPA of 3.83.
Has Sallar published research?
Yes. His peer-reviewed paper "Autonomous Road Defects Segmentation Using Transformer-Based Deep Learning Models With Custom Dataset" was published in IEEE Access, Volume 13, pages 174177–174199 (2025). DOI: 10.1109/ACCESS.2025.3617336.
Is Sallar available for freelance or contract work?
Yes — Sallar is currently working with international clients across the US, Canada, the UK, the UAE, Brazil, and Australia on full-stack and AI engineering projects, and takes on freelance and consulting engagements via Upwork or direct contact through sallar-ba.me.
What notable products has Sallar built?
Testorim (testorim.com) — an AI-powered QA testing platform that runs plain-English browser tests via Playwright with self-healing element resolution. Alif Ujala (alifujala.ai) — an interactive Urdu-learning platform for children combining Next.js web and mobile apps. Plus published computer-vision research on transformer-based road-defect segmentation.
What programming languages and frameworks does Sallar use?
TypeScript, JavaScript, Python, C++, C# and SQL. Frameworks: Next.js, React, Node.js, Express, FastAPI, Flask, Django. AI / ML: PyTorch, TensorFlow, LangChain, Hugging Face, OpenAI API. Data: PostgreSQL, MongoDB, Redis, Supabase. Cloud: Docker, AWS, GCP, Vercel.
How can I contact Sallar?
Through his portfolio at https://www.sallar-ba.me, GitHub at github.com/sallar-ba, LinkedIn at linkedin.com/in/sallar-ba, or directly via the chat assistant on the homepage. He also accepts engagements via Upwork.
What certifications does Sallar hold?
Coursera-verified certifications including the Machine Learning Specialization (DeepLearning.AI), Python Specialization (University of Michigan), Cloud Computing (IBM), Git and GitHub (Google), and Problem Solving and Decision Making (UC Irvine).

Working together

Which LLM providers do you work with?
OpenAI, Anthropic and open-weight models. I put them behind one interface so you can switch providers or fall back between them with a one-line change. The retrieval, evaluation and error handling around the model are where the time goes.
From Generative AI
How do you know the answers are correct?
We write down thirty or so real questions and the answers you would accept, early. Every change is measured against that set before it ships. Without it you are tuning prompts by feel.
From Generative AI
What language do you build AI systems in?
Python. FastAPI for services, PyTorch for models, LangChain where it helps, OpenCV and YOLO for vision. The web front end, when there is one, is Next.js and TypeScript.
From Generative AI
Do you work with clients outside Pakistan?
Yes. Most of my clients are in the US, Canada, the UK, the UAE, Brazil and Australia. I work remotely, with written updates and calls scheduled in your working hours.
From Generative AI
Who owns the code?
You do. The repository is yours from the first commit, and I hand over documentation so your team can run and change it without me.
From Generative AI
Do you train custom models or use existing ones?
Both, and choosing is part of the job. Most problems are solved faster with a pretrained model and careful data handling. When a custom model is justified I have done it: the IEEE Access paper trained transformer models on a custom road-defect dataset.
From AI & ML
Can you deploy the model, or only train it?
I deploy it. A model in a notebook is not a feature. I handle serving, monitoring and the integration into your product.
From AI & ML
How do we get started?
Email me or message me on Upwork with a few lines about what you need. We have a scoping call. I send back a written plan and an estimate. Work starts when you approve it.
From AI & ML
What can be automated?
Anything with a repeatable input and output. Content pipelines, lead handling, reporting, syncing data between tools, document processing. If someone on your team does the same thing twice a week, it is a candidate.
From AI Automations
What happens when an automation fails at 3am?
It retries, then it alerts you, and it leaves a log you can read. I build every workflow that way because an automation that fails silently costs more than doing the task by hand.
From AI Automations
Can you do real-time vision on standard cameras?
Yes. I have built passenger counting and parking monitoring systems on plain camera feeds with YOLOv8 and OpenCV, and the IEEE Access paper covers transformer-based segmentation on a custom dataset.
From Data Science
What stack do you use for web applications?
Next.js and React on the front end. Python FastAPI or Node.js on the back end. PostgreSQL or MongoDB. Deployed to Vercel or AWS. I choose within that based on what the product needs.
From Full-Stack Development
Can you take over an existing codebase?
Yes. I start with a short audit of the code, dependencies and deployment, write down what I find, and agree the first changes with you before touching anything.
From Full-Stack Development
Do you handle App Store and Play Store release?
Yes, including in-app purchase setup, review guidelines, signing and release management. I have shipped apps to both stores.
From Mobile Development
Can the app share a backend with our web product?
That is how I build it by default. One API and one database serve both clients, so there is one place to fix a bug.
From Mobile Development
Can you reduce our cloud bill?
Usually. The first step is measuring where the money goes. Most overspend is idle resources, oversized instances and missing caching. I fix what the numbers show.
From DevOps & Cloud
What does an engagement look like?
A fixed scope: an architecture review, an AI feasibility assessment, or an audit of an existing system. You get a written document with findings and a prioritised plan, and I can help execute it.
From Technical Consulting
Will you tell us if AI is the wrong tool for our problem?
Yes. I have built enough AI features to know which problems it solves and which it makes worse. Sometimes the useful answer is a database index, not a model.
From Technical Consulting
Do you work on existing Java systems?
Yes. Improving an existing service is usually worth more than rewriting it. I make it testable and observable first, then change it in steps you can ship.
From Java Development

Work with me

Question not here?

Ask it directly. I reply to every enquiry myself.

Email meHire on UpworkAll services