Snowflake
Software Engineer · Cortex Agents
Improving the quality and reliability of Cortex Agents, an agentic AI platform powering enterprise products like Snowflake Intelligence.
Go · Python · Large Language Models · AI Agents
Hey, I'm
Mikołaj!
pronounced: MEE-koh-why
I’m a software engineer who loves building reliable software. Currently working on agentic infrastructure at Snowflake ❄️.
Scroll
About
I build Cortex Agents at Snowflake, the agentic systems powering Snowflake Intelligence. My focus is reliability, quality signals, and production readiness at enterprise scale.
My path spans backend systems, distributed infrastructure, and ML research, shaped by time at Google, Meta, Amazon, and Quadrature Capital. I care about systems that are both technically rigorous and practically useful.
As a first-gen student from a small town in Poland 🇵🇱, I'm grateful for the opportunities and mentors I've had. Always happy to pay that forward, so feel free to reach out about software, ML, or starting a career in tech.
Outside of work, I enjoy investing, video and board games, and taking walks or hiking.
Experience
Google
Software Engineer Intern · Sovereign Cloud
Backend infrastructure for Google Sovereign Cloud.
Go · gRPC · GCP · Linux
More details
- Shipped two projects ahead of schedule for the Sovereign Cloud team and received a full-time return offer.
- Built a Go framework for containerized gRPC services, quickly adopted by 5+ teams.
- Designed a release validation system (Go, GCP) that became a core part of the team's CI/CD pipeline.
Quadrature Capital
Platform Engineer Intern · Systematic Trading
Platform infrastructure for Quadrature, a proprietary algorithmic trading firm.
Kubernetes · ArgoCD · Docker · Python
More details
- Expanded the Kubernetes-based platform supporting the automated trading system.
- Contributed to multiple projects simultaneously, addressing challenges across different business areas.
Amazon
Software Development Engineer Intern · Supply Chain Forecasting
Data pipelines for supply chain forecasting models serving Amazon’s logistics network.
Python · AWS · SQL
More details
- Optimized execution time of a data processing pipeline for supply chain forecasting models by ~90%.
- Automated a data ingestion workflow, saving 10+ hours of manual work weekly using Python and AWS.
Meta
Production Engineer Intern · Scribe
Reliability infrastructure for Scribe, processing terabytes per second of logs across Meta.
Python · Distributed Systems · Reliability Engineering
More details
- Implemented a performance testing service for Scribe, a critical log server handling TB/s of data.
- Expanded CI/CD pipelines to stress-test Scribe against hardware failures and heavy-load scenarios (chaos engineering).
Education
MSc Computer Science
2023 – 2025
Jagiellonian University
Specialization: Machine Learning
Focused on ML systems, explainability, and efficient model design. Was actively involved in research through the Group of Machine Learning Research (GMUM).
More details
Grade: 5/5 Distinction (summa cum laude)
Awards:
- MSc researcher scholarship on the XAI Paint grant: Interpretable and Interactive Multimodal Retrieval in Drug Discovery (Foundation for Polish Science, EU-funded).
- Student scientific grant for neural network pruning research (Excellence Initiative, Research University program).
- Rector’s Scholarship for Outstanding Academic Achievements.
BSc Computer Science
2020 – 2023
Jagiellonian University
Built strong foundations in algorithms, mathematics, and operating systems. Member and sysadmin of the Computer Science Students’ Association (KSI UJ).
High School Diploma
2017 - 2020
Small Town High School
Humble beginnings at a small public school in a small town in Poland.
Selected Works
Research publications and open-source projects in machine learning, AI explainability (XAI), and chemistry.
One Shot vs. Iterative: Rethinking Pruning Strategies for Model Compression
A comparison of one-shot and iterative pruning, with practical guidance for model compression.
OMENN: One Matrix to Explain Neural Networks
Convert any neural network into an interpretable matrix multiplication, explaining the model’s decision.
ProtoQuant: Quantization of Prototypical Parts for General and Fine-Grained Image Classification
An explainable method for neural networks which grounds decision in prototype parts from quantization.
ReAgentAI
An AI agent for chemical tasks, with tool integrations like retrosynthesis engine or molecular visualizations.
Contact
Open to research collaboration, interesting engineering problems, and conversations about AI systems. If you're early in your career or exploring tech, I'd be glad to help. Just reach out.
Email contact🍕janumiko.com click to open