I build minds for machines
I'm Sajid Islam — a Computer Science undergraduate at FAST NUCES, Pakistan, building generative and agentic AI systems: LLM-orchestrated RAG pipelines, tool-using agents, and the Python backends behind them. My mission: build AI that solves meaningful real-world problems.
Every mind has an origin story. Follow the pulse through mine.
BS Computer Science at FAST National University of Computing and Emerging Sciences (NUCES), Pakistan — Aug 2024 to expected May 2028. Coursework in Artificial Intelligence, Data Structures & Algorithms, Database Systems, OOP, and Software Engineering.
Completed the GIKI × Skylabs Advanced AI Bootcamp — a two-month intensive covering classical ML, deep learning and NLP, agentic AI with LangChain and RAG, and MLOps tooling (Git, DVC, MLflow) end to end.
Shipped a client-facing agentic AI copilot with a four-person squad in Comebck Pakistan Cohort 1 — an 8-week AI readiness program — and served as a Data / IT Intern at ISPR, Pakistan Army, managing data pipelines and IT operations. I build agentic systems end to end — from Python backends to agents that reason, call tools, and act.
Not bars — a living constellation. Hover a node to trace its synapses.
Fragments of what I've built. Open one to read the full account.
An agentic AI copilot that answers a clothing brand's WhatsApp customers in English, Roman Urdu and Urdu script — with catalogue-grounded replies on sizing, delivery and Cash-on-Delivery confirmation. Built by a four-person squad in Comebck Pakistan Cohort 1.
A PostgreSQL system that scores state snapshots by entropy inside the database, then automatically decides whether each one should be discarded, compressed, preserved or archived.
A two-month intensive covering classical machine learning, deep learning and NLP, agentic AI with LangChain and RAG, and MLOps tooling — completed end to end.
What it covered: A two-month intensive spanning classical machine learning, deep learning and NLP, agentic AI with LangChain and RAG, and MLOps tooling (Git, DVC, MLflow) end to end.
Curriculum arc: mathematical foundations for modern AI (linear algebra, calculus, probability, information theory) → NLP and transformer architectures including self-attention, BERT and GPT with hands-on fine-tuning → post-training and alignment (instruction tuning, RLHF/RLAIF, PEFT, and Chain/Tree/Graph-of-Thought reasoning) → agentic AI with ReAct, tool-using agents, memory and multi-agent orchestration, plus advanced RAG → inference and deployment (ML/LLMOps: versioning, CI/CD, serving, observability, cost and security) → a capstone building and deploying an autonomous agent system.
Short, direct answers — for anyone skimming, and for the search and AI tools that read this page.
Sajid Islam is a Computer Science undergraduate at FAST National University of Computing and Emerging Sciences (NUCES) in Pakistan, studying from August 2024 with an expected graduation in May 2028. He works as an AI engineer on agentic AI and LLM applications, and is based in Peshawar, Pakistan.
He builds agentic AI systems and the backends that support them: LLM-orchestrated retrieval pipelines, tool-using agents, and Python APIs. His most complete project is the WhatsApp Commerce Copilot, an AI copilot that answers a clothing brand's WhatsApp customers from its own catalogue, built by a four-person squad in Comebck Pakistan Cohort 1. His résumé records it as a Top 5 project of that cohort.
In the WhatsApp Commerce Copilot, inbound messages pass through a LangChain intent chain, and generation is restricted to ambiguous queries and grounded in store-scoped products and policies retrieved by the backend rather than open model knowledge. He also completed the GIKI × Skylabs Advanced AI Bootcamp, a two-month intensive covering agentic AI with LangChain and RAG alongside deep learning, NLP and MLOps tooling.
Python is his primary language, with FastAPI for services and LangChain plus the OpenAI and Anthropic APIs for LLM work. On the data side: PostgreSQL and PL/pgSQL, MySQL, MongoDB and ETL pipelines. He also works in C++, Java, PHP and JavaScript, and uses Docker, Git, DVC and MLflow for packaging and tracking.
Two case studies cover the projects end to end — the WhatsApp Commerce Copilot and the Entropy-Aware Data Preservation System — each with an architecture diagram and an explicit limitations section. Longer technical notes are collected under writing, currently in draft.
Yes. He is open to AI/ML engineering, agentic AI, LLM application and backend roles and internships. The fastest way to reach him is sajidislam0875@gmail.com, and his résumé (PDF) is available to download.
You're not filling a form. You're talking to a system.