Learn one layer down
If I’m using an ORM, I want to see the query. If I’m using LangChain, I want to know what actually got retrieved and why. The abstraction is a convenience, not an excuse.
About
Two disciplines,on purpose.
That combination is deliberate. Most AI work fails not because the model is wrong, but because nobody engineered the system around it — the retrieval, the state, the failure paths, the interface a real person actually touches. I’d rather own that whole surface.
- Discipline
- AI Full Stack Engineer
- Based in
- Lahore, Pakistan · Remote
- Graduating
- Jun 2027 (expected)
I’m a Computer Science undergraduate at the University of Management & Technology, Lahore, and I’ve spent the last two years doing the same thing in two different directions: shipping full-stack products, and teaching machines to act inside them.
- Education
- B.S. Computer Science
- University of Management & Technology (UMT) · CGPA 3.55 / 4.00
- Leading
- Project Lead, Fraternity
- 3–5 person team · Fall 2023 – present
Data Structures & AlgorithmsDatabase Management SystemsArtificial IntelligenceSystems Programming
01The Journey
From courseworkto production systems.
Five chapters, in order. Each one exists because the previous one exposed something I couldn’t do yet.
- Oct 202301 / 05
Started a CS degree and a team in the same semester
I enrolled at the University of Management & Technology in Lahore and, that same first semester, took on Project Lead at Fraternity — running a 3–5 person team through real project lifecycles. Leading before I felt qualified turned out to be the fastest way to find out what I actually didn’t know.
B.S. Computer ScienceProject Lead - Coursework02 / 05
Built the foundations properly instead of skipping to frameworks
Data Structures & Algorithms, Database Management Systems, Artificial Intelligence, Systems Programming — in C++, Java and SQL. I built small, finishable things alongside it: a reservation system on SQL Server, an Android notes app, a VLSM network design. Unglamorous, and the reason later work didn’t feel like magic.
DSADBMSSystems ProgrammingCGPA 3.55 - Logical Creations03 / 05
Learned what 'production' actually costs
As a Full Stack Developer Intern I shipped features across Next.js, Node.js and MongoDB — REST APIs, authentication, deployment, maintenance. The lesson wasn’t the stack. It was that auth, environment config and deploys are the job, not the paperwork after it.
Next.jsNode.jsMongoDBAuthREST APIs - Optaex Technologies04 / 05
Moved from calling models to engineering agents
At Optaex I worked on GenAI and Agentic AI: building and optimising agent workflows with LangChain, LangGraph and OpenAI APIs, RAG pipelines over vector databases, prompt engineering, evaluation, and getting intelligent agents deployed rather than demoed. This is where the two halves of my work met.
LangGraphLangChainRAGVector DBsEvaluation - Now → Jun 202705 / 05
Closing the gap between 'it works' and 'it scales'
I’m deliberately working on AWS, system design, vector databases and advanced agentic architectures while I finish the degree. The pattern I’m following hasn’t changed: learn the layer below the one I’m using, and ship something with it before moving on.
AWSSystem DesignAgentic AI
02Learning Philosophy
How I actuallyget better at this.
Anything I can’t explain simply, I don’t understand yet. Leading a team taught me that faster than any course did.
Finish small things
A finished small project teaches more than an abandoned ambitious one. Deployment, auth and edge cases only show up when you actually have to ship.
Rebuild to understand
The second implementation is where the learning is. The first one proves it’s possible; the second one proves I know why it works.
Explain it to keep it
Mentoring peers through programming and networking coursework forced precision. Teaching is compression, and compression is comprehension.
03Leadership
Responsibility taught memore than tutorials did.
Two roles at UMT, both of which meant being accountable for other people’s outcomes rather than only my own.
Project Lead
Fraternity, UMT
Fall 2023 – Present
Led a 3–5 member team through full software project lifecycles, organized tech meetups and coding competitions, and mentored peers in programming, networking, and core CS coursework.
Vice President
Anti-Narcotics Society, UMT
Organized awareness campaigns and student engagement initiatives in coordination with university administration and external stakeholders.
04Values
Four things I don’tcompromise on.
01
Ship it, then prove it
A feature isn’t done when it runs on my machine. It’s done when it’s deployed, authenticated, and someone else can break it without me watching.
02
Understand the layer below
I don’t want to only call the API. I want to know what the retrieval step actually returned and why the agent chose that tool.
03
Teach to consolidate
Leading a team and mentoring peers forced me to explain things precisely. Anything I can’t explain simply, I don’t understand yet.
04
Product before cleverness
The best architecture is the one a user never has to think about. Complexity is a cost, not a credential.
Currently learning
Next on the list.
AWSSystem DesignVector DatabasesAdvanced Agentic AI Architectures
Certifications so far: Claude 101 & AI Fluency (Anthropic) · Cybersecurity & Networking Basics (Cisco Networking Academy) · AI Essentials (Google) · Machine Learning (Coursera) · SQL (IBM).