COMPUTER VISION
Power Transformer Vision
Segmentation models that read a transformer assembly like a blueprint — catching mis-aligned components before they leave the floor.
YOLOv8m-segSAM2ONNXAzureCI/CD
// build log→
DOSSIER / 2026 · CASE FILE OPEN
Engineering
intelligence for
machines that
don't forgive errors.
Akash K Venugopalan — AI & Data Science engineer at Hitachi Energy, shipping computer vision and time-series systems for power transformers, steel mills and the messy real-world problems in between.
YRS
4+
FACTORIES
5+
MODELS SHIPPED
15+
PAPERS
04
- ROLE
- R&D · Data Scientist
- ORG
- Hitachi Energy
- EDU
- M.Tech AI
- LANG
- EN · HI · ML · TA
§01
Profile
I work where AI meets heavy industry — turning cameras, sensors and signals into decisions a plant manager can act on.
An M.Tech in Artificial Intelligence from Amrita Vishwa Vidyapeetham and a B.Tech in Electrical & Electronics Engineering from Saintgits frame how I think — equal parts data and physics. Today that means computer vision on transformer assembly lines, anomaly detection in steel mills, and the occasional LLM detour. Verified internationally by WES, Canada (2023).
| LOCATION | Kerala, India |
| TIMEZONE | UTC+05:30 |
| STATUS | Open to collab |
| DEGREE | M.Tech AI · 2022 |
| UNDERGRAD | B.Tech EEE · 2020 |
| CRED | WES Verified |
§02
Deployments
- 2024.12 — PRESENTSupports Sweden R&D · GlobalACTIVEHITACHI ENERGY
R&D Engineer · Artificial Intelligence & Automation
- Spearheading feasibility studies and developing cutting-edge computer vision and AI solutions from concept to deployment, including research, brainstorming, and PoC development.
- Driving Industry 4.0 transformation by integrating advanced AI technologies into Power Transformer production systems and industrial workflows.
- Leading the prototyping, validation, and optimization of AI algorithms to ensure reliable performance in real-world industrial environments.
- Designing and supporting CI/CD pipelines for seamless AI model deployment, scalability, and integration across factory ecosystems.
- Collaborating globally with research teams, academic institutions, and external AI partners to accelerate innovation and evaluate emerging AI technologies for enterprise adoption.
- 2022.07 — 2024.11Bengaluru, INARCHIVEDTOSHIBA R&D
AI Research Engineer
- Siamese vision models classifying anomalies captured by industrial cameras in steel rolling mills.
- Forecasted faults in Hot Strip Mills with ARIMAX / VARMA / FB-Prophet — strong RMSE and forecast-bias scores.
- Time-series anomaly detection with CNN, LSTM, GRU and attention models.
- Explainable AI for root-cause analysis of faults in time-series process data from steel mills.
- LLM research: Hugging Face T5 fine-tuning and prompt-tuning to generate code comments from source.
- 2021.09 — 2022.06Bengaluru, INARCHIVEDTOSHIBA R&D
AI Research Intern
- Anomaly detection with CNN-LSTM autoencoders and clustering methods (DBSCAN, K-Means).
§03
Systems
TIME SERIES
EQDS Forecasting Tool
Real-time Streamlit app visualizing FB-Prophet forecasts from the Equipment Quality Diagnosis System with interactive controls.
FB-ProphetStreamlitPython
// build log→
OPTIMIZATION
DT Optimization POC
AI-driven Distribution Transformer optimizer wired to a Bayesian optimization loop to push manufacturing efficiency.
Bayesian OptPythonScikit-learn
// build log→
LLM / NLP
Code-Comment LLM
Fine-tuned Hugging Face T5 with prompt-tuning to generate human-quality code comments directly from source.
Hugging FaceT5Prompt-tuning
// build log→
§04
Stack
Models & Methods08
- Computer Vision
- Instance Segmentation
- Siamese Networks
- LLM / NLP
- Time Series
- Anomaly Detection
- Explainable AI
- Bayesian Optimization
Languages & Frameworks08
- Python
- R
- TensorFlow
- Keras
- PyTorch
- Scikit-learn
- Hugging Face
- ONNX
Apps & Visualization04
- Streamlit
- Flask
- Matplotlib
- Seaborn
Cloud & MLOps05
- Azure
- Databricks
- Spark
- CI/CD
- Automation
§05
Licenses & Certifications
§06
Papers
| # | Title | Venue | Type | Year |
|---|---|---|---|---|
| 01 | Keypoint-Based Detection and Region Growing-Based Localization of Copy-Move Forgery in Digital Images↗ read | Springer · CVMI · IIIT Ahmedabad | CONFERENCE | 2023 |
| 02 | Copy-move Forgery Detection — A Survey↗ read | IEEE · ICICICT | SURVEY | 2022 |
| 03 | Study of Quantum Annealing and the Type of Related Applications↗ read | IEEE · PuneCon · MIT ADT | SURVEY | 2022 |
| 04 | AI-Based Audio Recognition System for Visually and Audibly Challenged↗ read | IEEE · ICOSEC | CONFERENCE | 2020 |
§07
Signal
$ ./contact --status open --rate > /var/log/inbox