Sacha
Lachin

Previously Junior Data Scientist (Mar. 2025 — May 2026)

• Member of the Data Science team for Theia, our Maritime Domain Awareness product that fuses together 20 million of square kilometres of satellite imagery with over 500 million AIS messages each day, providing invaluable insights to clients in both the government and commercial sectors
• Created state of the art ML models to detect irregularities in AIS (vessel transponder) data, as well as serving as an internal expert on this subject matter and working on modules to identify common behaviour patterns such as ship-to-ship transfers
• Won an internal competition to produce the most accurate model for predicting the future path of a vessel, all in a lightweight pipeline that returns an output in just a few seconds by leveraging lane graphs, physics-based modelling, DuckDB caches of vessel history, and LLM parsing of AIS messages
• Completely re-designed architecture of internal tools to halve the time spent by analysts reviewing events, significantly improving load times and simultaneously reducing database usage by means of caching and pre-loading large volumes of data
• Taken a leading role in the company in making use of agentic AI to increase workload efficiency, build and rapidly iterate prototypes, identify code inefficiencies and vulnerabilities, and conduct extensive pre-development research
• Deployed a large array of modules on our GCP cloud infrastructure, making use of schedulers, pub/subs, cloud runs, Terraform, and other technologies to implement multiple elements of the Theia platform that makes use of hundreds of microservices
• Took on the role as liaison between technical and intelligence collection teams, serving as the first point of contact for ingesting new data sources to the Theia platform, facilitating rapid iteration and deployment
• Worked cross-timezone with an international team spanning multiple continents to meet deadlines, fix bugs, and deliver proof-of-concepts rapidly

Previously Data Science Intern (Jul. 2022 — Sep. 2022)

• Developed and deployed client-facing LLM applications using Azure OpenAI, FastAPI, LangChain, and RAG architectures, including prompt engineering, solution design, and technical documentation
• Built backend and deployment infrastructure using Docker, CI/CD pipelines, Redis, PostgreSQL, SQLModel, and SQLAlchemy migrations, implementing safeguards to reduce operational and reputational risk in generative AI systems
• Designed data pipelines and evaluation frameworks for LLM-based tools using MLflow and Streamlit to quantify performance improvements for senior stakeholders
• Applied statistical and machine learning techniques including survival analysis, bootstrap resampling, and XGBoost to predict asset deterioration and building condition outcomes under sparse-data constraints
• Built Python-based data quality and geospatial processing tools, including anomaly detection, dataset scoring, reprojection pipelines, unit testing, and parallelised workflows to improve scalability and reliability

• Research Project: designed and proposed the Lachin-Stolarczyk method as an improved approach for the automated tuning of PID controllers in robotics systems, as part of a rigorous evaluation of heuristic and rule-based methods
• Undertook research by wrangling and visualizing clinical data whilst constructing machine learning models to predict sepsis in hospital patients
• Authored a case study on innovation and ethics in machine learning within the CAPTCHA industry
• Studied optional modules in data science, robotics systems, financial technology, and cloud computing