About
Hi, I’m Rahul. I’m a machine learning engineer at ACV Auctions, based in Chennai, India. I build production ML systems end to end: data pipelines, model selection and training, evaluation metrics, and efficient inference. I studied Computational Science and Engineering at the Technical University of Munich, where my master thesis was on few-shot object detection with CLIP.
I write here about machine learning, computer vision and LLMs. Posts are also published on Substack and Medium.
Reach me at psrahul.ml(at)gmail.com, or on GitHub and LinkedIn.
Experience
Machine Learning Engineer — ACV Auctions
October 2024 – present
Senior Software Engineer, ML Frameworks Team — MulticoreWare
May 2024 – October 2024
- Machine learning inference for Texas Instruments TDA4 chips.
Computer Vision and Machine Learning Developer — TÜV Rheinland AG
May 2023 – April 2024 · Bochum, Germany
- Ran statistical analysis on 20k+ data samples and built end-to-end data pipelines with MongoDB, FastAPI and Label Studio for streamlined data retrieval.
- Designed, trained and evaluated PyTorch models for vehicle damage detection. Introduced a new class-balancing algorithm that raised precision from 40% to 70%.
- Deployed the models as microservices with TorchServe REST APIs on a Kubernetes cluster, cutting inference time from 272 ms to 41 ms per sample.
Machine Learning Researcher (Master Thesis) — PreciTaste
May 2022 – December 2022 · Munich, Germany
- Proposed CLIPtheCenter, a multimodal few-shot object detection model that combines OpenAI’s CLIP with the CenterNet architecture.
- Raised object detection precision from 17% to 50% over the CenterNet baseline.
Machine Learning Engineer (Part Time) — PreciTaste
October 2021 – April 2022 · Munich, Germany
- Designed and implemented data-driven vision algorithms in PyTorch to segment regions of interest in videos, replacing hand-written rules.
- Trained and deployed object detection models with up to 95% mean precision for production projects, and coordinated data annotation and model maintenance for a customer deployment.
- Developed a web app with Plotly, Dash and Pandas for visualising experimental data.
Machine Learning Intern, Video Analytics Team — IPS Intelligent Video Software, Securiton Germany
July 2021 – September 2021 · Munich, Germany
- Developed deep learning proofs of concept for detecting pedestrians in security camera footage with TensorFlow and OpenVINO.
- Managed automated annotation of 5000+ images and set up SQL data pipelines to retrieve data by object category.
- Ran experiments with models such as the TensorFlow Object Detection API on a GPU cluster with parallel training, and designed metrics to balance inference speed against accuracy.
Research Assistant, Computer Vision Team — Technical University of Munich
- Contributed to dataset preparation for PhoCaL, published at CVPR 2022.
Student Teaching Assistant — Technical University of Munich, Introduction to Deep Learning (I2DL)
November 2020 – March 2021
- Held office hours and helped prepare programming exercises.
Graduate Engineering Trainee — Ashok Leyland
July 2017 – August 2018 · Chennai, India
- Worked on predictive reliability analytics for automotive components, measuring product reliability from warranty claims and spare parts data.
- Standardised onboarding of lateral hires as a functional Six Sigma project in Human Resources.
- Managed relationships with axle and brake drum suppliers during periods of high product demand.
Education
M.Sc. Computational Science and Engineering — Technical University of Munich
October 2019 – December 2022
GPA 1.9 (German scale). Coursework included Machine Learning, Introduction to Deep Learning, Advanced Topics in 3D Computer Vision, Visual Data Analytics, Data Innovation Lab, Advanced Programming, Parallel Programming, and the seminar Deep Learning for Medical Applications.
M.Tech. Computational and Experimental Mechanics — Indian Institute of Technology Madras
Partly completed; specialised in fluid dynamics and computational methods.
B.Tech. Mechanical Engineering — SASTRA Deemed University
June 2013 – May 2017
GPA 1.4 (German scale). Coursework included Numerical and Statistical Methods, Programming in C, and Programming in C++.
Skills
- Languages and databases: Python, C++, MongoDB, Redis
- Machine learning: PyTorch, TensorFlow, scikit-learn, Optuna
- MLOps: MLflow, Kubeflow, TorchServe, Label Studio, FastAPI
- Numerical: NumPy, SciPy, OpenCV, scikit-image, Pandas
- Developer tools: Kubernetes, Docker, Earthly
- Generative AI: OpenAI API, LangChain, Hugging Face, diffusion models
Languages
English (fluent) · Tamil (fluent) · German (intermediate, Goethe-Zertifikat B1)
Publications
- PhoCaL: A Multi-Modal Dataset for Category-Level Object Pose Estimation with Photometrically Challenging Objects. P. Wang, H. Jung, Y. Li, S. Shen, R. P. Srikanth, L. Garattoni, S. Meier, N. Navab, B. Busam. CVPR 2022. arXiv — contributed to dataset preparation.
- CLIPtheCenter. Master thesis on few-shot object detection, Technical University of Munich, 2022. Project · Report
- Synthetic Learning: Learn From Distributed Asynchronized Discriminator GAN Without Sharing Medical Image Data. Seminar article, Deep Learning for Medical Applications, TUM. Read
Achievements
- Dean’s Merit List scholar for top 2% academic performance (B.Tech., SASTRA University).
- University topper in Numerical and Statistical Methods and in Programming in C++ (B.Tech., SASTRA University).
- First rank in the April 2014 semester examinations.
Certifications
- Languages: Goethe-Zertifikat B1 (German)
- Cloud: AWS Certified Cloud Practitioner
- Specializations: Deep Learning · AI for Medicine
- LLMs: Open Source LLMOps Solutions
- Generative AI: Build Basic GANs · Build Better GANs
- Machine learning: Structuring Machine Learning Projects · The Data Scientist’s Toolbox · Neural Networks and Deep Learning · Introduction to TensorFlow for AI, ML, and DL · Convolutional Neural Networks in TensorFlow · Improving Deep Neural Networks · Sequence Models
- AI for medicine: Medical Diagnosis · Medical Prognosis · Medical Treatment