Data architecture, AI reliability and business meaning.
I'm Dmitry Ustimov, a Data & AI Architect with 15+ years building data platforms, analytics systems and ML infrastructure, and a former Head of Data.
I've led data organizations in fintech, built systems at 100TB to petabyte scale, and worked across analytics, governance, real-time services and executive decision systems, as both an engineer and a leader.
Today I focus Decision Spine on one problem: making AI analytics reliable enough to use for real business decisions. I build the research in public through the AI Analytics Harness and Preflight, then apply those findings to real data environments.
Fifteen years of data systems that moved real numbers, across fintech, e-commerce, telecom and healthcare.
12%
customer-base growth in 8 months
A company-wide metric tree at Zeal Group surfaced declines early and pinpointed the cause, so leadership could act — the kind of decision system this practice builds.
< 30 ms
p99 ML anti-fraud service
At Coins.ph, the service that secured the company's operating license and held its customer base through fraud-heavy Covid.
+10%
customer retention
A production ML recommendation engine at < 100 ms p99, built on a ~400TB data lake at Lamoda.
+35%
deposit success rate
Part of a Zeal Group data strategy across trading, anti-fraud, compliance and product — pricing stability improved 8% alongside.
100TB–PB
data at scale
Platforms for BI, executive decisions, fraud and compliance — from ~200TB at Coins.ph to petabyte-scale ETL at MegaFon.
25
in the data organization
Built and led engineering, BI, data science, platform and infrastructure teams from inception across Coins.ph and Zeal Group.
300+
self-service BI dashboards
A Metabase self-service platform at Coins.ph, with 90% of financial reconciliation and reporting pipelines automated.
PII/PHI
healthcare data governance
At Neko Health, privacy and governance readiness for sensitive multi-country health data — data classification, PII/PHI tagging, least-privilege access, audit logging and cataloging on Databricks Unity Catalog.
- Decision Spine
The commercial practice: benchmark, diagnose and repair AI analytics reliability in real environments.
The open-source research lab: controlled experiments on what makes an AI analyst reliable.
Open-source static analysis that catches ambiguous analytics definitions before an agent runs.
Outside Decision Spine, Dmitry also builds HabiBubble, an AI-native geospatial intelligence product.