Skip to content
View buchbindern's full-sized avatar

Block or report buchbindern

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
buchbindern/README.md

Hi, I'm Natalie

Applied ML engineer working on predictive maintenance and reliability — Bayesian state estimation and survival analysis, plus the systems that run them in production.

Founded Kaful, building prognostic digital twins for industrial equipment. Open to hands-on engineering roles in industrial AI and manufacturing.

Portfolio · LinkedIn

Projects

kaful-data-twin — Streaming CNC digital twin, deployed as multi-tenant SaaS. Estimates tool wear cut-by-cut from raw cutting-force data with a particle filter and forecasts remaining useful life. Session auth, per-tenant isolation, concurrent ingest (~2× throughput, with the remaining ceilings documented). Postgres + R2 on Render. Live demo →

Main finding: cross-tool RUL transfer fails because of the observation model — the wear→force map — not the degradation rate. Isolated with a 2×2 study, reproducible from the repo.

Kaful agentic pipeline — Compiles OEM manuals into executable physics-based digital twins (RAG + LLM extraction → simulation → Monte Carlo RUL) with no historical failure data. Validated across four machines. Presented at IWSM 2026.

Battery_RUL_CNN — PyTorch CNN predicting lithium-ion battery RUL from single-cycle discharge data. R² ≈ 0.76 on held-out cells.

SurvivalSimulation — Out-of-sample comparison of survival models (Cox PH, DeepSurv, RSF, GBSA) under varying censoring.

Tech

Python · PyTorch · PostgreSQL · TypeScript · Next.js

Particle filtering · survival analysis · agentic LLM/RAG pipelines · streaming ingest & concurrency · multi-tenant services

Pinned Loading

  1. SurvivalSimulation SurvivalSimulation Public

    Out of sample comparison of survival models (Cox PH, RSF, Gradient Boosting) under varying censoring rates.

    Python

  2. battery-rul-cnn battery-rul-cnn Public

    PyTorch CNN for predicting lithium ion battery remaining useful life from cycle discharge data.

    Python

  3. kaful-showcase kaful-showcase Public

    Physics grounded prognostic digital twins for industrial equipment from OEM documentation — no historical failure data required.

    HTML