I help growing businesses and teams figure out where AI creates real value, then architect and integrate the systems that actually deliver it — reliably, on time, and built for production.
I work close to the problem — not just the code. Each engagement starts with understanding your domain before anything is built.
Your problem is mine. I dig into your workflows, data, and pain points to map where AI genuinely helps — and where it doesn't. You get a clear picture before a single line of code is written.
Already have a product or platform? I integrate LLMs, agentic workflows, and RAG pipelines directly into your existing stack with secure APIs, minimal disruption, and proper observability.
From architecture through to production deployment: agentic workflows, multi-agent orchestration, model fine-tuning, and cloud-hosted APIs built to be maintainable and scalable as your needs grow.
Every project starts with the business problem, not the technology.
I immerse myself in your domain — your workflows, data sources, and pain points. We define what success looks like before designing any solution.
I architect the agent pipelines, data flows, and integration touchpoints. You see a clear plan — including security, compliance, and where cloud infrastructure will be needed — before any code is written.
Production-grade code: secure APIs, RBAC, observability, and tests. Deployed on AWS or GCP, containerised with Docker, with CI/CD pipelines so iteration is fast.
AI systems need ongoing calibration. I work with your domain experts post-launch to monitor outputs, catch edge cases, and keep improving the system as real-world data comes in.
An end-to-end agentic workflow that automates patient triage — from structured symptom intake through AI-assisted pre-diagnosis, doctor recommendation, and intelligent appointment booking — while keeping a physician in the loop at every critical decision point.
A selection of recent builds — prototypes, research projects, and production experiments.
Takes natural language input and generates detailed floor plans using the FLUX-Schnell diffusion model. Includes a raster-to-vector conversion module and a layout quality evaluation function — outputs are usable in architectural visualisation and game development.
Flask app that lets you ask questions about any YouTube video in multiple languages. Embeds transcripts with FAISS and Hugging Face multilingual embeddings, and answers via a Gemini-powered RAG pipeline with maintained session context.
Research project comparing LSTM and GRU architectures for expanding short ~50-word user outlines into coherent full-length stories. Trained on scraped and cleaned large-scale text data. GRU outperformed LSTM on short sequences due to faster convergence.
Tools I've worked with hands-on, in real projects — not just tutorials.
I'm Asadullah Yousaf — an AI Engineer based in Lahore, Pakistan, with a couple of years of hands-on experience building agentic and RAG-based systems that do real work in production.
My approach starts with the business domain, not the technology. I spend time understanding your workflows, your data, and what "done" actually looks like before designing anything. Most of my work involves integrating AI into existing systems — helping teams that already have a product get meaningful AI capabilities without rebuilding from scratch.
My strongest areas are agentic workflow design, RAG pipelines, and LLM integration via secure APIs. I've worked across healthcare and EdTech, and I'm comfortable deploying and managing AI workloads on AWS and GCP. I'm still learning — every production system teaches me something — but I'm rigorous about getting things right before they go live.
Let's find out together. I'm happy to talk through your problem before we discuss any engagement.