Skip to content
View anson10's full-sized avatar
🎯
Focusing
🎯
Focusing
15 followers · 7 following

Achievements

Achievement: StarstruckAchievement: Pull Sharkx2Achievement: YOLOAchievement: Quickdraw

Achievements

Achievement: StarstruckAchievement: Pull Sharkx2Achievement: YOLOAchievement: Quickdraw

Highlights

  • Pro

Block or report anson10

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.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. 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
anson10/README.md

Anson Antony

ML, data and real-time systems for semiconductor manufacturing
M.Sc. Computational Materials Science, TU Bergakademie Freiberg

Website • Email

About Me

  • M.Sc. student in Computational Materials Science at TU Bergakademie Freiberg (since Oct 2025)
  • B.E. Mechanical Engineering, Anna University, Chennai (2024)
  • Based in Freiberg, Germany. English C1, German B1, Tamil native
  • Open to Werkstudent and HiWi roles in ML, data engineering, MLOps, backend Python and simulation, in Saxony or remote across Germany

Projects

Three connected projects take one problem from model to factory floor. Every result is scored against ground truth or a chance baseline.

Project What it does Key results Stack
FabEye Wafer-map failure pattern recognition on real fab data (WM-811K) 0.858 macro-F1 on a lot-disjoint split; conformal review rule auto-accepts 96.2% of wafers at a 1.9% error rate on unseen lots PyTorch, ONNX Runtime, FastAPI, Docker
WaferLens Yield-excursion detection and root cause on a simulated fab with logged ground truth Chamber EWMA/CUSUM alarm after a median 8–9 affected wafers vs. 33 for Shewhart; SQL commonality analysis ranks the true chamber first for 72% of yield-loss excursions (chance: 0.8%) PostgreSQL/TimescaleDB, dbt, Dagster, Kafka, Power BI
WaferEdge Real-time wafer-map watchdog at the equipment, speaking SEMI HSMS/SECS-II Sort result to lot-hold decision in 0.3–0.5 ms median at 1,000–5,000 wafers/s; own CUDA engine runs FabEye's CNN at ~11,000–13,000 maps/s C++23, CUDA, AVX2, Asio, CMake

Also: PhaseNet, phase identification from LAMMPS molecular dynamics snapshots, trained with PyTorch DDP. QM9-GPT, inverse molecular design on QM9.

Tools I Use

Languages :     Python, C++23, CUDA, SQL, Bash
ML        :     PyTorch (DDP), PyTorch Geometric, scikit-learn, LightGBM, ONNX Runtime, MLflow
Data      :     PostgreSQL/TimescaleDB, dbt, Dagster, Kafka (Redpanda), FastAPI, SQLAlchemy
Systems   :     CMake, AVX2, lock-free queues, ASan/TSan, libFuzzer
DevOps/BI :     Docker Compose, GitHub Actions, Prometheus, Grafana, Power BI, Linux/HPC
Simulation:     LAMMPS, OVITO

Pinned Loading

  1. waferLens waferLens Public

    Fab yield excursion detection and root cause on a simulated fab with ground truth: SPC benchmarked against injected excursions, commonality analysis, FabEye wafer-map patterns, real-time SPC on Kaf…

    Python

  2. WaferEdge WaferEdge Public

    In-line wafer-map watchdog at the equipment edge: SECS/GEM, AVX2, CUDA, own CNN inference engine (C++23)

    C++

  3. FabEye FabEye Public

    Wafer map defect pattern classification on WM-811K: lot-disjoint evaluation, conformal uncertainty, and an ONNX/FastAPI service

    Python 1

  4. Path-Finding-Visualiser Path-Finding-Visualiser Public

    An interactive tool built with C++ and SFML to visualize algorithms like BFS, DFS, A*, Dijkstra, and Greedy Best-First Search.

    C++ 7

  5. PhaseNet PhaseNet Public

    High-Performance Deep Learning Framework for Thermodynamic Phase Identification from Molecular Dynamics Simulations

    Python

  6. QM9-GPT QM9-GPT Public

    QM9-GPT is a deep learning platform that performs Inverse Molecular Design process of starting with a desired physical property and generating the chemical structure that satisfies it.

    Jupyter Notebook 1