Name: Soumyabroto Banerjee
Current Role: Area Manager, AI
Current Organization: Tata Steel Ltd
Experience:
Tenure: 2020 - (till date)
Email: soumyabroto.banerjee@gmail.com
Phone: (+91) 94775-55030
LinkedIn: soumyabroto-banerjee
GitHub: SOUMYABROTOBANERJEE
Skills
Agentic AI : 80%About Me
Area Manager, AI at Tata Steel Ltd. Leading digital transformation with GenAI, Agentic AI, and Data Science. Recognized for building Tata Steel Digital Assistant (flagship GenAI platform), Code Genie (enterprise coding assistant), and Sarathi (voice-based business travel assistant). Winner and finalist at Tata Innovista, published in TSL Innovator Journal. Passionate about automation, storytelling, and impactful presentations.
Previously at TCS Digital (AI Engineer), with experience in NLP chatbots, MRI classifiers, and Qlik Sense visualizations. B.Tech in Electronics & Communication from University of Calcutta (TATA Millennium Scholar, IEEE STEX Runner-up).
Interests: GenAI, Agentic AI, Deep Learning, Computer Vision, NLP, Business Analytics, Teaching, Drama, and Photography.
Education:
B.Tech, Electronics & Communication, University of Calcutta (2016-2020), CGPA: 9.12
TATA Millennium Scholar, IEEE STEX 2018 Runner-up
Don Bosco School, Kolkata (ISC: 94.5%, ICSE: 91.8%)
Services
The creme-de-la-creme of what I can achieve in a deadline.
Computer Vision Developer
Working with Python on Deep Learning Projects - Both Research and Applications based. I ace at Computer Vision - Core Areas - Related to Humans - Facial, Emotional, Age, Gender, Gaze, etc. Other Areas of interest - Psychological Tracking, Bio-Medical Image Processing, Classical Filtering techniques, etc.
NLP Analyst
Take keen interest in Proper Representation of Psychological Patterns in texts. Chatbot and Conversational Analytics Developer at TCS - Building a Chatbot to connect to external sources like Qlik, Tableau, etc. also able to deliver ad hoc augmented analysis on the fly. Interacts like human, Functions like a Machine.
Business Analyst (Machine Learning)
Started learning about the Learning of Machines in my 2nd Year of College. Specialty in Supervised - Working on end to end Suppy Chain Customer Bucket Problems, Understanding Time Series and its impact on Business, Understanding the Inventory, and predicting the Order Patterns in time.
Machine Learning Researcher
Worked on Clustering Problems - Identifying weak & hard clustering ways - Generating and ensembling to make softer predictions of multiple weak learners, - Moving of Cluster Points in a way/pattern to make cluster shape and its index look better in an iterative fashion.
GenAI & Agentic AI SME
Leading GenAI adoption at Tata Steel. Architect of Tata Steel Digital Assistant (LangChain, Google ADK, LangGraph), Code Genie (enterprise coding assistant), and Sarathi (voice-based business travel assistant). Experience with Google Palm2, Imagen, Azure GPT-4, DallE. Solutions for structured, unstructured, and semi-structured data. Private Beta tester for Google GenAI.
Teaching Guide
"If you want to master something, Teach it !" - Feynman. I absolutely love teaching. So far, I have managed to teach Python and basic Deep Learning. My area of expertise encompasses - Teaching in TCS, Subject matter Expert at Chegg to Teaching Kids and underprivileged children. Recently, added Gen AI to this list too.
Portfolio
Showcase of Completed and Compiled Works
Mask Based Facial Expression Recognition
Adben Industries | AWS Cloud | Real-time facial recognition with anti-spoofing and mask detection. Used in attendance systems. Optimized with TensorRT. GitHub
Multimodal Brain Tumor Detection
Springer Best Paper Award. CNN-based classifier merging MRI+SPECT using SWT/PCA. Published in AMPHE 2020. Read the paper
MOCHA (Man Operated Car Having AI)
IoT and AI based Car controlled by a basic DTMF. It also streams a live video feed with labeled objects in its view. Runs on AlexNet and via a phone call - Connect anytime, anywhere and get your tasks missed done through a Robotic Hand. Read the documentation here
Cluster Learner
'n' Hard clustering algorithms are made to pass through a voting mechanism. The hard clustering are first made soft by a probabilistic equation & modification of points lead to better clustering shapes and increased probabilities of outlier detection. Awaits Publication.
BioMedical Brain Tumor Classification
Multimodal Brain Tumour Classification with RGB SWT Blended Images. Built a custom CNN Classifier for the task. Results compared with VGG-16 and Augmentation. Published - Awaiting Chapter. Read the documentation here