Dionizije Fa

My way of doing things.

I’ve learned the rarest advantage is simple: caring deeply about the work. I approach work and life with the now famous Elon's principles:

Job Experience

Software Engineer / Founding Member — Entropic

OmicsAgent agent and infra platform for bioinformatics.

  • Built OmicsAgent, an AI agent system that runs bioinformatics workflows on cloud compute.
  • Designed and implemented the core backend for agent orchestration: tool calling, workflow execution, telemetry, and logging around long-running bioinformatics jobs.
  • Built evaluation datasets and an open-source benchmarking framework for testing LLMs on bioinformatics tasks (code generation, tool use, interpretation of genomics results).
  • Implemented MCP (Model Context Protocol) scaffolding so OmicsAgent tools can be exposed cleanly to multiple LLM clients and frontends.
  • Worked directly with B2B customers (research labs / biotech) on integration: mapping their existing pipelines to our platform, sizing compute tiers, and adapting the product to real workloads.
  • Designed and maintained the Azure architecture (storage, compute, networking), including cost modeling and budgeting for long-running analysis jobs.
  • Stack Python (FastAPI, orchestration, agents), Svelte (frontend), Go (CLI/tooling), Azure (Container Apps / Batch / Storage), Docker, Micromamba, Snakemake-style workflows, Postgres, OpenTelemetry / logging, GitHub Actions / CI.
2024 — Present
Python Team Lead — Digacon Software Solutions

StrömungsRaum AI platform for fluid dynamics simulations.

  • Worked on StrömungsRaum an AI platform for fluid dynamics simulations for industrial chemistry.
  • Led a remote team of 4 outsourced developers: hiring, technical interviews, onboarding, mentoring, code reviews, etc.
  • Worked full stack on a platform for running AI-enabled simulations on customer geometries; delivered production features and models for large German industrial clients, mostly for extrusion processes.
  • Implemented and optimized geometry/mesh algorithms (cleanup, feature extraction, mathematical and ML modeling for geometry optimization, boundary condition tooling) in Python/C++.
  • Worked on features for predicting simulation results from initial conditions.
  • Implemented a GPU-accelerated visualization of simulation results in the web viewer.
  • Worked on and maintained deployment Kube/Docker workflows for deployment.
  • Stack Python (scientific stack, mesh/geometry tooling), C++ (core algorithms, performance-critical parts), Vue.js (frontend), REST APIs, GPU-accelerated visualization (PyVista/WebGL-style pipelines), Docker, Hetzner VMs, Linux, GitLab CI.
2023 — 2025
Software Engineer / Founding Member — Sanitas Analytica

Polypharm Solutions FHIR API for personalized and safe drug prescribing.

  • Helped build Polypharm Solutions a FHIR API for personalized drug prescribing in hospitals. a clinical decision support system focusing on drug safety, drug–drug interactions, side effects, and pharmacogenomics.
  • Designed and implemented the backend FHIR API in Python/Flask, including data models for patients, medications, genotypes, and clinical rules.
  • Implemented modules for drug interaction checking, drug metabolism and enzyme saturation, linking pharmacogenomic markers to dosing recommendations, and side-effect risk analysis on patient cohorts.
  • Used statistical methods (propensity score matching, basket analysis, cohort analysis) to relate predicted metabolism saturation to real-world side-effect patterns (e.g. by gender, drug group).
  • Built the Angular frontend used by clinicians for patient-level views, prescribing support, and cohort reports.
  • Deployed and maintained the system on AWS (app servers, Postgres, backups, monitoring) and handled releases and troubleshooting in production.
  • Product reached ~1000 active users across medical universities, private practices, and hospitals before being replaced by a government solution.
  • Stack Python (Flask), Angular, Postgres, AWS, FHIR APIs, REST, statistical modeling (Python / R), Docker, CI/CD, Git.
2021 — 2023
Research Assistant — Ruđer Bošković Institute

PhD research in applied ML in Machine Learning and Knowledge Representation lab.

  • Worked as a research assistant on applied ML projects in genomics, chemistry, and protein-protein interaction networks.
  • Built GNN and attention-based models in PyTorch/PyTorch Geometric for representation learning on biological graphs (e.g. PPI networks). Example: CCL-PPI Graph Attention models (ISMB/ECCB 2021 Representation Learning in Biology).
  • Implemented classical ML pipelines in scikit-learn, plus statistical analysis in R / statsmodels for lab collaborators.
  • Did a lot of genomics work using Bioconductor and related R tooling (differential expression, enrichment, basic pipeline glue).
  • Helped other researchers by running analyses, cleaning data, and turning one-off scripts into reusable notebooks and small libraries.
  • Wrote and contributed to external grant proposals and projects (e.g. AI4EU Drug-Attrition-Oracle), handling both the technical plan and early prototypes.
  • Deployed experimental services on local GPU machines and small Kubernetes clusters using Docker and gRPC for internal APIs.
  • Presented work at conferences and internal seminars, explaining ML methods to mixed audiences (biologists, chemists, computer scientists).
  • Stack Python (PyTorch, PyTorch Geometric, scikit-learn), R (Bioconductor, statsmodels-equivalent in R), Jupyter, Docker, Kubernetes (basic usage), gRPC, Git, Linux, basic HPC/GPU workflows.
2020 — 2021

Publications & Talks