Open to opportunities

Milos Plavsic

@milosplavsic

Message

AI and Data Science leader building production ML systems that scale.

Zimbabwe

Message

What I'm looking for

I’m looking for a role where I combine technical depth with team leadership—hands on engineering, strong technical direction, and the ability to translate business needs into scalable execution in a more technical environment.

I’m a quantitative researcher/analyst and engineering leader with a PhD and deep experience building production data science systems. I’ve led teams, architected high performance data pipelines, and delivered machine learning solutions that work at scale.

At HeldaWay.ai, I co-founded an institutional-grade financial transaction database and analytics API, built and reviewed models, and coached a cross-functional team of up to 8 engineers and data scientists. I also re-engineered core pipelines from Pandas to Polars to cut latency and memory usage, and engineered relative value and term-structure/bond pricing systems for decision support.

Experience

Work history, roles, and key accomplishments

Founder and CEO

Improvement Institute

Jan 2022 - Jan 2026 (4 years)

Founded an alternative approach to pre-university education and led the community and company operations. Implemented AI agent-based solutions using Ollama with Llama 3 and Llama 4 models to support applied AI initiatives.

AI Agents Ollama LLaMA Generative AI Python Applied Machine Learning

Co-Founder, Head of AI & DS

HeldaWay.ai

Jan 2018 - Jan 2022 (4 years)

Co-founded an institutional-grade financial transaction database and analytics API, leading AI/data science direction and client discovery. Re-engineered core pipelines from Pandas to Polars, reducing request latency up to 15x and memory usage by 50%, and built relative value screening for mispriced bonds using PCA and clustering.

Clustering Polars Python Financial Time Series API Development

Quantitative Analyst

Rule Financial

Jan 2007 - Jan 2011 (4 years)

Developed and optimized ETF tracking error models and built a no-arbitrage yield curve model incorporating liquidity premiums and collateral convenience yields. Implemented an OIS discounting engine to transition valuation from LIBOR to collateral-aware curves, reducing swap valuation errors by 15 basis points.

Fixed Income & Rates Modeling No Arbitrage Yield Curves OIS Discounting ETF Tracking Error Modeling Model Validation Python SQL