Open to opportunities

Leylani Mugambi

@leylanimugambi

Message

AI Engineer — LLM Applications · RAG Pipelines · Conversational AI · Healthcare AI

Kenya

Message

What I'm looking for

Open to: Junior/Mid AI Engineer · LLM Developer · Conversational AI · Prompt Engineer · AI Data Annotator · Remote · Contract · Full-time

I build AI-powered products that work in production.

My most recent build: MediAssist Kenya (live at huggingface.co/spaces/Purityk/mediassist) — an AI clinical decision support system built with Python, LangChain, OpenAI API, and Streamlit. Clinicians can ask it anything about drug protocols, treatment guidelines, and diagnostics. It's live, deployed, and getting real queries.

I also built HerCycle, a reproductive health AI app for East African women — combining cycle tracking with an AI symptom checker and maternal health guidance.

What I bring:
→ End-to-end AI engineering: from data ingestion to deployed product
→ RAG pipeline architecture using LangChain, vector embeddings, and OpenAI API
→ Conversational AI with multi-turn memory and context management
→ Prompt engineering for safety-critical healthcare contexts
→ Optimisation for low-bandwidth, mobile-first African environments

Tech stack: Python · LangChain · OpenAI API · Streamlit · FastAPI · Pinecone · Hugging Face · Git/GitHub · Docker

I'm a Clinical Officer by background — which means I understand the domain problems that AI needs to solve, not just the code. That makes me uniquely effective at building healthcare AI that actual clinicians trust and use.

Open to: Junior/Mid AI Engineer · LLM Developer · Conversational AI · Prompt Engineer · AI Data Annotator · Remote · Contract · Full-time

📍 Nairobi, Kenya — available globally, all timezones

🔗 GitHub: github.com/pkirimi
🔗 Live Demo: huggingface.co/spaces/Purityk/mediassist
🔗 LinkedIn: linkedin.com/in/purity-karwitha-9357a41b6
📧 purity.kirimi.k@gmail.com

Experience

Work history, roles, and key accomplishments

Current

AI Clinical Decision Support

Current

MediAssist

Jan 2024 - Present (2 years 5 months)

Built a RAG-powered clinical decision support assistant that lets clinicians query medical guidelines, drug databases, and diagnostic protocols in natural language, reducing lookup time by an estimated 70%. Engineered an end-to-end pipeline (document ingestion, embeddings, retrieval, and LLM response generation) with citation tracking and optimized latency/token usage for 2G/3G environments.

RAG Data Pipelines OpenAI API Pinecone Embeddings Data Ingestion Retrieval Chains fastAPI LLM Citations

remote

hercycle

May 2025 - Apr 2026 (11 months)

I built HerCycle, a reproductive health AI app for East African women, combining cycle tracking with an AI symptom checker and maternal health guidance.

Remote AI AI Tools Lang Chain Language Models Pinecone Streamlit Python

Clinical Officer & Telehealth

Penda Health

Jan 2021 - Jan 2023 (2 years)

Worked on AI-driven healthcare initiatives by validating clinical AI outputs, refining model responses, and contributing domain expertise to AI-assisted diagnostic tools. Optimized digital health workflows, increasing patient retention by 30%, and delivered remote consultations with 85% NPS.

Telemedicine Medical Case Review Clinical Documentation RLHF Adjacent Evaluation Patient Retention NPS Driven Service Improvement

Clinical Officer Intern

Kenyatta National Hospital

Jan 2019 - Jan 2020 (1 year)

Supported diabetes research and telehealth advancement efforts with hands-on exposure to data collection, documentation, and digital health systems. Contributed to early-stage work that improved readiness for research and technology-enabled care delivery.

Clinical Research Support Telehealth Support Data Collection Medical Documentation Digital Health Case Documentation Early Healthcare Data Handling

Reproductive Health AI Chatbot

Jan 2024 - Present (2 years 5 months)

Developed a conversational maternal and reproductive health chatbot providing guidance, symptom triage, and education for underserved East African users. Applied prompt engineering and multi-turn context/memory to keep responses clinically safe, culturally appropriate, and coherent across sensitive topics.

Conversational AI Prompt Engineering Streamlit Healthcare NLP Culturally Safe Responses