Joseph Mwangi

Software Engineer

Sasabot • Full-time

Nov 2025 - Present • 9 mos

Architected and scaled the core Agentic AI infrastructure, successfully transitioning a vulnerable MVP into a highly reliable, production-grade automated chat platform.​ ●​ Designed and deployed multi-agent frameworks leveraging Model Context Protocol (MCP) and advanced vector search, drastically reducing LLM latency and optimizing context retrieval accuracy.​ ●​ Engineered high-throughput REST APIs and webhooks using Python (FastAPI) and Go, implementing robust caching layers with Redis and MongoDB to handle concurrent real-time user traffic​ . ●​ Integrated secure financial infrastructure, including M-Pesa APIs, to automate transactional workflows, payment processing, and real-time chat triggers.​ ●​ Established strict CI/CD pipelines and production monitoring, utilizing memory profiling and optimization techniques to eliminate critical server bottlenecks and memory spikes.​ ●​ Collaborated directly with the CTO and CEO to align engineering velocity with the product roadmap, while leading technical onboarding and code reviews for incoming engineering talent.

Machine Learning & MLOps Engineer​

Brain Technologies • Freelance

Jan 2024 - Dec 2024 • 11 mos

Developed and deployed multiple LLM-powered RAG chatbots for technical customer support using LangChain, Hugging Face, and vector databases (FAISS, Pinecone). ●​ Fine-tuned transformer-based models (LLaMA 2, Mistral) for domain-specific tasks, improving response accuracy by 35%. ●​ Integrated OpenAI APIs for multi-turn conversation handling and sentiment-aware responses. ●​ Built internal tooling for prompt engineering, retriever tuning, and evaluation, leading to a 25% reduction in hallucinations and a 40% boost in relevance scores. ●​ Designed CI/CD pipelines for model deployment using FastAPI, Docker, and GitHub Actions; ensured99.9% uptime. ●​ Collaborated with cross-functional teams to translate business objectives into scalable LLM workflows.