Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

Hi, I'm Iman Khatibi — an AI and Machine Learning Engineer with 3+ years of hands-on experience designing transformer-based systems, RAG pipelines, and evaluation-driven model optimization. I specialise in modular RAG architectures, fine-tuning LLMs, and improving grounding, latency, and cost efficiency in generative systems through structured experimentation. I thrive on building robust tooling and pipelines that make AI systems reliable, safe, and scalable. I enjoy turning messy data into actionable insights and deploying production-grade AI solutions that solve real-world problems, from healthcare chatbots to data analytics platforms.…Hi, I'm Iman Khatibi — an AI and Machine Learning Engineer with 3+ years of hands-on experience designing transformer-based systems, RAG pipelines, and evaluation-driven model optimization. I specialise in modular RAG architectures, fine-tuning LLMs, and improving grounding, latency, and cost efficiency in generative systems through structured experimentation. I thrive on building robust tooling and pipelines that make AI systems reliable, safe, and scalable. I enjoy turning messy data into actionable insights and deploying production-grade AI solutions that solve real-world problems, from healthcare chatbots to data analytics platforms.

Iman Khatibi

Data Scientist, AI Engineer, Full Stack Developer, +3





Hi, I'm Iman Khatibi — an AI and Machine Learning Engineer with 3+ years of hands-on experience designing transformer-based systems, RAG pipelines, and evaluation-driven model optimization. I specialise in modular RAG architectures, fine-tuning LLMs, and improving grounding, latency, and cost efficiency in generative systems through structured experimentation. I thrive on building robust tooling and pipelines that make AI systems reliable, safe, and scalable. I enjoy turning messy data into actionable insights and deploying production-grade AI solutions that solve real-world problems, from healthcare chatbots to data analytics platforms.…Hi, I'm Iman Khatibi — an AI and Machine Learning Engineer with 3+ years of hands-on experience designing transformer-based systems, RAG pipelines, and evaluation-driven model optimization. I specialise in modular RAG architectures, fine-tuning LLMs, and improving grounding, latency, and cost efficiency in generative systems through structured experimentation. I thrive on building robust tooling and pipelines that make AI systems reliable, safe, and scalable. I enjoy turning messy data into actionable insights and deploying production-grade AI solutions that solve real-world problems, from healthcare chatbots to data analytics platforms.

Available to hire

Hi, I’m Iman Khatibi — an AI and Machine Learning Engineer with 3+ years of hands-on experience designing transformer-based systems, RAG pipelines, and evaluation-driven model optimization. I specialise in modular RAG architectures, fine-tuning LLMs, and improving grounding, latency, and cost efficiency in generative systems through structured experimentation.

I thrive on building robust tooling and pipelines that make AI systems reliable, safe, and scalable. I enjoy turning messy data into actionable insights and deploying production-grade AI solutions that solve real-world problems, from healthcare chatbots to data analytics platforms.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Language

English

Fluent

Work Experience

LLM Engineer Intern at HoFin

August 1, 2024 - December 1, 2024

Designed and implemented a production-oriented mental health chatbot leveraging retrieval-augmented generation (RAG) over curated domain documents, deployed on AWS using EC2 and S3 for document storage and inference workflows. Architected modular RAG pipeline separating ingestion, chunking, embedding, vector indexing, retrieval, and response generation layers. Improved grounded response relevance by around 20% (Precision@K) through systematic tuning of chunk size, overlap, embedding model selection, and top-K thresholds. Reduced hallucination frequency in multi-turn conversations by 15% by enforcing retrieval-first generation and integrating conversational memory control. Decreased token usage per response by 12% through structured context compression and prompt refinement. Experimented with embedding and inference endpoints on Azure OpenAI and GCP Vertex AI for performance benchmarking. Built evaluation framework combining automated semantic similarity scoring and structured human rev

Data Scientist researcher at University of Iceland

January 1, 2023 - January 1, 2025

Fine-tuned FLAN-T5 and BERT models on domain datasets using LoRA and mixed precision training. Improved ROUGE-L score by 9% compared to base FLAN-T5 after parameter-efficient fine-tuning. Reduced training time by 25% through mixed precision and optimized batch scheduling. Automated hyperparameter search with Optuna, reducing manual tuning cycles by around 30%. Conducted controlled experiments comparing decoding strategies (beam search, temperature scaling, length penalties) to optimize fluency vs factual alignment. Evaluated summarization and classification outputs using ROUGE, F1, and cross-validation.

Data Analyst at HoFin

February 1, 2022 - August 1, 2024

Built data transformation and reporting pipelines supporting operational and product strategy. Analyzed behavioral and performance datasets to identify optimization opportunities.

AI Engineer at HoFin

August 1, 2024 - November 1, 2025

Designed and implemented a production-oriented mental health chatbot leveraging retrieval-augmented generation (RAG) over curated domain documents, deployed on AWS using EC2 and S3 for document storage and inference workflows; Developed a decision-making AI agent that analyzes user messages and dynamically routes conversations to LLM responses, RAG retrieval, or human escalation using LangChain and LangGraph; Implemented core agent capabilities including tool calling, conversational state management, safety classification, and multi-step reasoning; Architected modular RAG pipeline separating ingestion, chunking, embedding, vector indexing, retrieval, and response generation layers; Improved grounded response relevance by around 20% (Precision@K) through systematic tuning of chunk size, overlap, embedding model selection, and top-K thresholds; Reduced hallucination frequency in multi-turn conversations by 15% by enforcing retrieval-first generation and integrating conversational memory

Data Analyst

February 1, 2022 - August 1, 2024

Built data transformation and reporting pipelines supporting operational and product strategy. Analyzed behavioral and performance datasets to identify optimization opportunities.