# Baljinnyam Dayan

> Forward Deployed Engineer at PhysicsX in London. MSc Artificial Intelligence from Imperial College London. Previously machine learning and data engineering at Rio Tinto.

- Website: https://www.balj.in/
- Email: hi@balj.in
- GitHub: https://github.com/baljinnyamday
- LinkedIn: https://www.linkedin.com/in/bdayan
- Google Scholar: https://scholar.google.com/citations?user=YKqED58AAAAJ
- X: https://twitter.com/baljinnyamday
- Location: London, United Kingdom

## Work

### PhysicsX, Forward Deployed Engineer (September 2026 to present)

London. https://www.physicsx.ai

### Rio Tinto, Data Systems Engineer, Data & ML Engineering (May 2024 to September 2025)

Worked on the Oyu Tolgoi copper mine, open pit and underground. https://www.riotinto.com

- Automated manual rock stability inspections underground with a fault-tolerant computer vision pipeline (SAM and YOLO, fine-tuned for mining imagery) that processed 1,000+ images an hour and saved $145K a year in engineer hours. Went underground with the geotechnical engineers to learn what makes a drawpoint safe: rock size distribution, wetness and wall condition.
- Built a real-time rainfall monitoring pipeline that cut false-positive mine shutdown alerts by 30%, preventing 12+ unnecessary production halts.
- Deployed a tailings dam stability model across 8 facilities, with 95% of site engineers using it for daily risk review. Training was orchestrated with Apache Airflow and MLflow.
- Built a resource governance system in Palantir Foundry that found and decommissioned 2,400+ stale datasets, saving $120K a year in compute.

### dentsu, Software Engineer, Generative AI (December 2022 to April 2024)

https://www.dentsu.com

- Built an end-to-end generative AI system that turns multimodal input into print-ready label and packaging artwork.
- Optimised GCP deployment and CI/CD for generative AI applications.

## Projects

### NerdReels (2026 to present)

AI educational video generation. https://nerdreels.com

- A multi-agent pipeline that handles research, scene planning, visual QA, narration and animation from a single prompt. A one-minute video renders in about three minutes.
- Asynchronous rendering backend on FastAPI, Celery, Redis, AWS SQS, Docker and S3 with CloudFront delivery.

### Arbitration.mn (2022 to 2024)

Co-founder and engineering lead. A secure legal arbitration platform for the Mongolian International Arbitration Center. https://www.arbitration.mn

- National digital signatures, SSO and strict access control over sensitive case data.
- Worked alongside 15+ judges and legal professionals across 3 jurisdictions to digitise case workflows, cutting processing time from weeks to days.

### Bagsh.Space (2022 to 2025)

Course planning and professor reviews for university students. https://bagsh.space

- An academic recommendation system used by 40K+ students, with 78% retention.
- Integrated with the academic data APIs of 3 universities.

## Education

### Imperial College London, MSc Artificial Intelligence (2025 to 2026)

- Dissertation: Agentic AI for Network Attack Simulation and Red Teaming.
- Y Combinator AI Startup School. Palantir Hackathon, 1st place. Imperial and Bezos Centre for Sustainable Protein Hackathon, runner-up.

### National University of Mongolia, BEng Software Engineering (2020 to 2024)

Distinction, GPA 3.9/4.0.

## Research

- Lung Disease Detection with Vision Transformers (2024). https://arxiv.org/abs/2411.11376

## Skills

- AI and ML: PyTorch, computer vision, YOLO, SAM, LLMs, agents, LangGraph, conformal prediction
- Engineering: Python, TypeScript, Next.js, FastAPI, Celery, Redis, Kafka, Spark, Docker, Palantir Foundry, AWS, GCP, CI/CD