Looking for a job

Sayan Chaudhuri

@sayanchaudhuri

Data scientist and ML engineer specializing in production ML, physics-informed deep learning, and real-time spectroscopy.

India

Message

What I'm looking for

Seeking a challenging role in applied ML/AI where I can build production-grade deep learning systems, work on physics-informed modeling or computer vision problems, and grow within a research-driven, collaborative engineering team.

I am a Data Scientist and ML Engineer with ~2.5 years of industry experience building and deploying production machine learning systems across spectroscopy, EV thermal modeling, and NLP pipelines.

My M.Tech from IISc Bangalore focused on computer vision and domain generalization; I improved Mask R-CNN with custom augmentation and regularization, achieving a 7.6% accuracy gain on corrupted benchmarks.

At Picarro, I replaced a 15s classical optimization with a 10ms 1D Vision Transformer for infrared spectral analysis, generated 200M synthetic spectral mixtures, and reduced training time from 26 hours to 2.5 hours while establishing MLflow-based monitoring and KPIs.

Previously at Mercedes-Benz R&D I built Azure ML production pipelines for EV battery thermal prediction, developed physics-informed losses and VAE-based anomaly detection, co-invented two filed global patents, and shipped RAG pipelines and FastAPI services for production use.

Experience

Work history, roles, and key accomplishments

Current

Data Scientist

Current

Picarro

Aug 2025 - Present (6 months)

Replaced slow physics-based optimization with deep learning achieving faster inference (15s → 10ms) for real-time infrared gas analysis. Improved detection accuracy from 78% to 84% using a 1D Vision Transformer with physics-informed constraints. Generated 200M synthetic spectra, trained distributed models on AWS, and productionized MLflow-based tracking, monitoring, and validation pipelines.

PyTorch MLFlow Signal Processing Synthetic Data Generation Jax Python Deep Learning Machine Learning Transformers AWS Sagemaker Vision Transformer Representation Learning Model Evaluation Experiment Design Time Series Analysis Distributed Training Technical Presentations

Junior Data Scientist

Mercedes-Benz R&D India

Aug 2023 - Jul 2025 (1 year 11 months)

Designed and deployed end-to-end ML pipelines on Azure for EV battery thermal prediction using physics-informed deep learning trained on synthetic and real sensor data. Built VAE-based anomaly detection and production FastAPI services for fleet-scale inference. Developed RAG pipelines using GPT and FAISS, and co-invented two patented ML-based thermal optimization systems.

Databricks MLFlow fastAPI Faiss Machine Learning Deep Learning Variational Auto Encoders Anomaly Detection Regression Modeling Physics Informed Modeling Cloud Computing Machine Learning Pipelines Microsoft Azure REST APIs RAG Large Language Models Synthetic Data Generation LangChain Libraries

Education

Degrees, certifications, and relevant coursework

Indian Institute of Science

Master of Technology, Aerospace Engineering

2021 - 2023

Grade: 8.4/10

Activities and societies: Coursework in Computing for AI & ML, Numerical Optimization, ML for Signal Processing, Reinforcement Learning, Robotic Perception.

Completed M.Tech in Aerospace Engineering with thesis on robust object detection and domain generalization improving Mask R-CNN performance on corrupted benchmarks.