Lead Backend Developer at Pharos Production
Dmytro Ihnatov documents backend architecture and R&D experiments from the team's work on a professional learning platform for aesthetic medicine. He is the author of the development materials behind this series.
Backend engineering and R&D
His work on the platform covers retrieval pipelines, source processing and backend services. The articles trace specific decisions through their constraints and recorded results. They also identify where a test, implementation plan or incomplete experiment leaves a question open.
From the development record
The series follows a learning platform from text retrieval to figures and recorded media. Cases examine citation support, transcription errors and the state changes needed to keep derived material inspectable. Project and book names are anonymized where required; the articles distinguish reported experiments from fictional explanatory examples.
Articles by Dmytro Ihnatov
Accurate OCR, Wrong Answer: Evaluating Medical Tables
All 33 checked table rows had exact cells, but a live answer used the wrong device column despite good page-level OCR and retrieval scores
What a Low WER Missed in a Medical Transcript
Our synthetic interview scored about 1.4% WER, while a separate FDA recording contained four FAIRS substitutions that the attention rules did not flag
When a Vision Model Turns Weeks into Months
Two readings flagged a unit error in our six-figure spike, while the later 22-figure run showed why numeric disagreement remained a limited inspection aid
What Our RAG Test Doubles and Synthetic Corpora Could Actually Prove
A 31.3-second recording reported a 43.1-second bookmark, so our test generator needed checking before it could judge model timing and retrieval behavior
Valid Citations, Unsupported Claims: What Fifteen Questions Revealed About Our RAG
A patient-specific answer passed our citation parser while exceeding its sources, exposing the gap between source markers and claim support in a September test
How technical review is assigned
AI articles are assigned to Victor Sineglazov. FinTech, crypto, blockchain and regulatory topics are assigned to Olena Zaichenko. Dmytro Nasyrov reviews other engineering topics. Completed review is credited on the relevant article with its actual review date. The first five articles were technically reviewed by Victor Sineglazov, with completion confirmed on October 15, 2026.
Last updated:
