I build AIthat ships.
I'm an AI/ML engineer shipping LLM systems, agentic RAG, and voice agents for teams from London to California. Two years in, with open-source work 2,300+ developers have starred, currently pursuing an MSc in AI at LUMS.
Research to
production.
I'm an AI/ML engineer. I take LLM research and turn it into production systems — RAG pipelines, fine-tuned models, and agents that survive real traffic. Two years in, I've shipped for teams in London, California, and Lahore. One of my open-source repos has 2,300+ stars. I'm currently pursuing an MSc in AI at LUMS.
I don't chase benchmarks. I chase what survives real users, real traffic, and 3 AM incidents. Every project below was built end-to-end and deployed — not left in a notebook.
Things I've built.
Case studies — not screenshots. Each project below includes the problem, the approach, and the measurable outcome.
CureWise — Agentic RAG for Healthcare
Problem
Doctors and patients needed one interface to parse medical reports, diagnose skin conditions from images, and book appointments — without the hallucinations that kill clinical trust.
Impact
Zero-hallucination responses grounded in real-time PostgreSQL patient history. Fine-tuned vision models diagnose acne and rashes from uploads. Agents hit a calendar API to book appointments directly from chat.
- LangChain
- LangGraph
- Pinecone
- Docker
- PostgreSQL
- GPT-3.5
LangChain RAG — Grounded QA at Scale
Problem
GPT models hallucinate when asked about proprietary documents. Teams needed a way to get grounded answers over their own corpora without retraining.
Impact
Built a semantic QA pipeline with Hugging Face embeddings and FAISS retrieval feeding GPT-3.5-turbo. Answers cite source chunks. Drop-in architecture for any custom document set.
- LangChain
- FAISS
- HF Embeddings
- OpenAI
- Python
Automated Summary Evaluator (FYP)
Problem
Educators scoring ADHD student summaries by hand faced consistency issues and long turnaround times. NGOs needed an automated way to grade content and wording at scale.
Impact
Trained LLMs as regressors on curated ADHD datasets from multiple NGOs. Shipped via GitHub CI/CD and Docker for reproducible deployment. Graded summaries in seconds instead of minutes.
- Python
- LLMs
- CI/CD
- Docker
- GitHub Actions
Built in public.
Five years of building in the open. 2,400+ stars and 475+ forks across my repos — developers learning from and building on my work.
Beginner-Data-Science-Projects
A curated collection of hands-on data science projects for beginners — the most starred repo in my portfolio.
Curated collection of books and learning resources covering Deep Learning, LLMs, and Generative AI.
Computer vision project for gesture recognition using OpenCV, built for sign language and gesture-based interfaces.
TensorFlow-based image classifier handling varying numbers of classes with a flexible CNN architecture.
Deep learning face detection pipeline using OpenCV with data augmentation and neural networks.
Production-grade healthcare platform using LLMs, RAG, and agentic workflows for medical analytics and conversational AI.
Where I've been.
Datality
- Designed intelligent dashboards using LangChain agents and LLMs, empowering data-informed decisions with enhanced accuracy.
- Engineered ML pipelines with TensorFlow automating insight generation and predictive analytics for business intelligence.
- Stack: APIs, NLP, LLMs, Docker, AWS, TensorFlow, GitHub.
Qult Technologies
- Fine-tuned Llama 3.1 8B for NLU tasks, achieving a 20% accuracy gain on text classification use cases.
- Automated end-to-end ML pipelines on AWS SageMaker + Lambda with OpenAI tools for scalable deployment.
- Stack: APIs, NLP, LLMs, Docker, AWS, TensorFlow, GitHub.
BornGreat
- Built sentiment analysis over social media data using NLP + LLMs to surface supply, demand, and competitor signals.
- Shipped a FastAPI app for real-time model interaction, wired to GitHub Actions CI/CD for zero-touch deploys.
- Stack: Selenium, APIs, NLP, LLMs, FastAPI, Docker, TensorFlow.
Spyresync
- Built interactive dashboards and visualizations over cleaned, engineered datasets inside Django.
- Integrated ML models into web apps to provide predictive features.
- Stack: Python, APIs, GitHub, Django, Scikit-learn.
The stack
I ship with.
Tools I reach for when something has to work in production. Chosen for reliability, debuggability, and what actually survives a 2 AM page.
- Python
- C / C++
- SQL
- LangChain
- LangGraph
- LlamaIndex
- OpenAI APIs
- Hugging Face
- TensorFlow
- PyTorch
- Keras
- Scikit-Learn
- NLTK
- OpenCV
- Docker
- AWS
- FastAPI
- GitHub
- Pinecone
- Selenium
- Jupyter
Education & awards.
Have an AI
problem worth solving?
I'm open to full-time AI/ML roles, contract work on LLM and RAG systems, and research collaborations. Fastest reply is email — I usually respond within 24 hours.
taimour.a.karim@gmail.com→