Maroš Jančo

Best Startup Award. AI clinical-documentation platform for therapists.

• Own prompts, models and quality for AI-generated progress notes (core product) across 250+ documentation sections; shipped 100+ production releases across multiple model generations without quality regression. Mentored new hires; reported AI roadmap and cost trade-offs directly to founders.
• Lowered AI costs by ~50% while improving quality. Built an LLM-as-judge eval framework on Langfuse (datasets, eval runs, trace-level flow debugging) plus an A/B-testing pipeline; used to gate releases, catch regressions and shorten the prompt-iteration cycle.
• Productionised RAG, prompt orchestration and agentic patterns across Gemini, Claude, GPT-* and Llama; benchmarked Vertex AI, AWS Bedrock, Azure OpenAI and Anthropic continuously, routing traffic by use case rather than committing to one provider.
• Automated a Claude-Agent-SDK customer-support agent (internal-tool integration, hardened prompts and guardrails).
• Ran LLM observability and MLOps (Langfuse, Grafana, BetterStack, AWS CloudWatch, SNS, SQS, Azure, Google Cloud, Vertex AI, Superset); lifted team velocity via Claude Code in daily workflow, improved CI/CD, and internal sessions on modern RAG / prompting / agentic best practices.

• Led ideation and planning of an expert-validated knowledge-graph platform; reshaped business model, scope, financials and technical architecture for a successful UK Smart Funding submission after 12 months of prior poor-quality work.

Global Impact 50 Award 2023. Physical climate risk for enterprise assets.

• Built from scratch the first physical-climate-impact platform assessing combined-hazard damages on assets and portfolios worldwide; shipped a financial-impact model predicting stock-price shocks from adverse climate events.
• Led a global wildfire-damage prediction product end-to-end and built a company → subsidiary → asset framework for SP500 and FTSE100 firms; optimised distributed pipelines over TBs of geospatial, satellite and financial data (PySpark, Databricks, Geopandas, Xarray, k8s).

Global investment-bank R&D lab.

• Built a history-augmented collaborative-filtering recommender for client document recommendations (Amazon S3, Spark, Dask); weekly chatbot recommendations save thousands of clients hours of search.
• Designed and shipped an NLP entity-extraction system from scratch — Random Forests, GBT, Kneser-Ney n-gram LMs, Naive Bayes and Word2Vec into one interpretable classifier; later scaled with LSTM- and BERT-based architectures. A dozen+ resulting NER models now automate chat-to-price FX execution worldwide daily.

Lexomat · lexomat.sk →

Solo-founded AI legal-research chat for Slovak and Czech law, indexing millions of documents; drove product, design, agentic AI flows, pricing, GDPR and infra end-to-end (one contractor on delivery). Stack: Next.js / TypeScript, Python / FastAPI, Supabase + PGroonga, Vertex AI.

Souli · github.com/dzino-app/souli.app →

Open-sourced end-to-end-encrypted AI companion that gamifies personal growth. Privacy-first architecture — chat content encrypted end-to-end, including in transit to and from the model. Full-stack TypeScript + privacy engineering.

PayToEat.sk

Co-founded and ran a food-delivery app (2019); led product, ops and growth across Slovak towns. Pre-AI era; the business still serves customers today in one Slovak town.