Saeid Masoumzadeh
Lead AI / Machine Learning Engineer · PhD in Computer Science
About
Versatile AI engineer specialising in recommendation systems, large language models, and agentic AI systems. Full-stack ML engineer who turns product goals into shipped features end-to-end — from discovery and design to implementation, evaluation, and cloud deployment (AWS / GCP). Strong focus on LLM efficiency, orchestration design, and scalable AI infrastructure.
Core Skills
Recommender Systems
Two-tower models, ANN/vector search, learning-to-rank, re-ranking, cold-start & session-based recommendation.
LLMs & Agents
RAG pipelines, tool-use / MCP (LangChain, LangGraph), prompt design, guardrails, caching strategies, agent control-plane design.
Product & Delivery
Stakeholder management, roadmap definition, experiment design, cross-functional leadership, client-facing consulting.
Data & Feature Eng.
SQL (BigQuery), Pandas, ETL/ELT, feature pipelines (Airflow, DVC, Metaflow), data quality monitoring.
MLOps & Platforms
Model serving, CI/CD, monitoring (Datadog), Vertex AI, Bedrock, SageMaker, vector search integration.
Cloud & Infra
GCP, AWS, Docker, serverless & batch compute, vector search (FAISS, Pinecone).
Experience
Senior AI Engineer
9fin · London, UK
- ›Designed and deployed an AI retrieval agent using LangGraph, Elasticsearch and AWS, improving precision@10 by 32% in a pilot evaluation of financial search queries.
- ›Built an end-to-end financial document processing pipeline combining MinerU and LLM agents, with automated statement drafting, validation and traceability to source pages and cells, deployed to staging.
- ›Established automated LLM-as-judge evaluation, human benchmarking and production monitoring to measure retrieval quality, identify failures and guide performance optimisation.
AI Engineer (Contract, R&D)
S&P Global · London, UK
- ›Optimising agentic workflows in LangChain/LangGraph — graph execution control, node parallelisation, selective state propagation.
- ›Integrating LiteLLM for semantic caching and unified LLM API gateway routing.
- ›Designing an internal LLM evaluation platform for benchmarking prompts, tool-use, and multi-step reasoning.
- ›MCP tool-calling via code-generation patterns to minimise token usage and improve determinism.
- ›Deploying agentic applications on AWS AgentCore for scalable LLM orchestration.
Lead Machine Learning Engineer
Motorway · London, UK
- ›Architected a multi-objective ranking framework for a two-sided marketplace — 4% revenue uplift.
- ›Launched an agentic AI Bidding Assistant enabling conversational interaction with recommendations.
- ›Built an agentic Condition Centre for real-time prep/repair cost estimation.
- ›Mentored a high-performing team; established rigorous experimentation practices.
Senior Data Scientist
Ankorstore · London, UK
- ›Led deep-learning retrieval to enhance product discovery using NLP and multimodal signals.
- ›Built a CLIP-based multimodal tagging system — 5% engagement uplift.
Senior Data Scientist
Lyst · London, UK
- ›Designed a two-tower deep model for item embeddings leveraging behavioural signals.
- ›Advanced discovery with ML, NLP & computer vision; ran rigorous A/B tests.
Founder & CTO
Cyra · London, UK
- ›Built a multi-agent AI assistant for recruiters sourcing top candidates via ML/NLP.
Entrepreneur in Residence
Entrepreneur First · London, UK
- ›Founded Cyra; secured initial funding; validated business model with EF mentors.
Education
PhD in Computer Science
University of Vienna, Austria
Thesis: Autonomic Management of Virtual Machines in Cloud Data Centres Using Machine Learning.
MSc in Artificial Intelligence
Azad University, Tehran, Iran
Thesis: Intelligent and Adaptive Active Queue Management in Packet Switching Networks.
BSc in Computer Engineering (Hardware)
Azad University, Tehran, Iran
Personal Projects
Customer Support AI Assistant
Flan-T5 fine-tuning with public Twitter data; combined in-context learning and PEFT; deployed a PoC API.
Chat With Your Candidate
Recruiter-facing chat over uploaded CVs using a quantised LLaMA-class model and LangChain.