Company
Celsus
10 cases from Celsus (Healthtech), each digested from a public talk or engineering post and linked to the original. 3 report results in production.
Tasks: Detect anomalies, Build a data platform, Reduce ML cost. Methods: Data augmentation, LoRA fine-tuning, Label noise detection, Human / expert labeling, LLM simulation / role-play, Synthetic data generation (LLM) and 22 more. Tools: PostgreSQL, Whisper, Kubernetes, Amazon S3, FastAPI, Grafana and 26 more.
A Synthetic Training Pipeline for Medical LLMs![]() | ![]() | healthtech | 2026-05 |
Challenges of Labeling Doctor-Patient Dialogues for ASR![]() | ![]() | healthtech | 2025-05 |
Diffusion Models for Synthesizing "Horrible" Medical Data![]() | ![]() | healthtech | 2025-05 |
Supervised Labeling: When the Rulebook Doesn't Work![]() | ![]() | healthtech | 2024-05 |
Incidents in ML Services and How to Monitor Them![]() | ![]() | healthtech | 2024-05 |
Data-Centric AI: A Review of Methods![]() | ![]() | healthtech | 2023-05 |
Selecting and monitoring physician annotators![]() | ![]() | healthtech | 2022-06 |
From JSONs to a Data Platform![]() | ![]() | healthtech | 2022-06 |
Active Learning![]() | ![]() | healthtech | 2022-06 |
A review of data cleaning and annotation quality control methods![]() | ![]() | healthtech | 2022-06 |
