Tushir Sahu

Product Engineer

Suita, Osaka, Japan0 mo experience

Key Highlights

Enhanced unit test coverage to 97% in MLOps.
Developed real-time traffic video processing pipeline.
Proposed innovative architectures for medical image segmentation.

Stackforce AI infers this person is a Machine Learning and Computer Vision specialist with a focus on SaaS and Healthcare applications.

Contact

Skills

Core Skills

MlopsMachine LearningComputer VisionDeep Learning

Other Skills

C++Data StructuresFastAPIGenerative Adversarial Networks (GANs)GrafanaKubernetesMLrunMathematicsObject DetectionProblem SolvingPrometheus.ioPythonReinforcement LearningUnit Testing

About

I like research :‑).

Experience

Ge vernova

DT Intern

Jan 2025 – Jul 2025 · 6 mos · Hyderabad, Telangana, India

Contributed to the development and maintenance of the Model Verification Manager (MVM) platform by enhancing unit test coverage to 97% across multiple modules, leading to improved code reliability and easier refactoring.
Designed and implemented a pipeline-level ExitHandler to ensure consistent failure detection of pods across the MLOps workflow, capturing and logging critical errors to reduce debugging time.
Improved error traceability and observability through structured and semantic logging practices, facilitating faster issue resolution and better incident reporting during pipeline executions.

MLOpsUnit TestingMachine LearningPythonDeep Learning

(株) vrain solution

AI Advanced Researcher

Nov 2024 – Dec 2024 · 1 mo · Tokyo, Japan · Remote

Worked on creating Fine-Tune prompt encoder for MobileSAMv2 to improve the performance of VRAIN product
Developed a training script for MobileSAMv2 incorporating the fine-tuned prompt encoder
Added a Multihead Attention mechanism along with FCN for box generation with refinement, replacing the use of YOLOv8-seg for box generation

Computer VisionMachine LearningPythonDeep Learning

Indian institute of technology, kanpur

Research Intern

May 2024 – Jul 2024 · 2 mos · Kanpur, Uttar Pradesh, India · On-site

Created an unsupervised synergistic double decoder integrating a pretrained model with a trainable decoder while selectively freezing layers.
Leveraged weight sharing by concatenating outputs from pretrained and trainable layers for improved efficiency.

Deep LearningMachine LearningPython

Kaspertech

Computer vision intern

Feb 2024 – May 2024 · 3 mos · Remote

Created a multi-threaded video processing pipeline for real-time traffic video processing.
Developed deployment strategies of the trained models and integrated the created models with set servers.
Worked on running machine learning models on continuous feeds whilst reducing the bandwidth and lag.

Computer VisionMachine LearningPython

Indian institute of technology, roorkee

Research Intern

Aug 2023 – Dec 2023 · 4 mos · Remote

Proposed two architectures SpeedNet and ConstructNet in the field of medical image segmentation and object detection respectively.
Worked on developing Involution-LSTM architecture for the task of Next frame video understanding.

Deep LearningMachine LearningPython