Jul 2022 – Nov 2023 · 1 yr 4 mos · Bengaluru, Karnataka, India

• Led the Data Science team of Embibe that works on the following areas:
• Generative AI -
• Academic Chatbots
• Question Answering using RAG and Chain of Thoughts reasoning
• Augmented Analytics - Report Generation from Data
• NLP - Semantic Search, Multilingual Search, Neural Machine Translation (NMT)
• Computer Vision - Image Detection, Image Similarity Search
• Intelligent Tutoring Systems - Bayesian Knowledge Tracing, Knowledge Graph
• ML Systems -
• Ingestion: Airflow for Ingestion Pipelines.
• Training: MLFlow for Model & Dataset versioning, Onnx for Model Quantization.
• Inference: FASTAPI for Model Serving, MLFLow for inference monitoring, NVidia Triton inference Server.
• Databases: PostgreSQL, MongoDB, ElasticSearch, Weaviate - Vector Database
• Key Results:
• Embibe Lens: Led the team to build the backend of Embibe Lens App that enables students to scan a book or paragraph using a smartphone's camera, select the highlighted concepts, and watch them come to life in high-definition 3D Science models.
• Semantic Search: Implementation of Semantic Search using Sentence Transformer (MiniLM) and Vector Database (Milvus) resulted in reduction of Null results by 12pp and increase of top-3 CTR by 4 pp from the Lexical Search.
• Chatbot: Implementation of Embibe Doubt Resolution chatbot for Indic Languages using Yolo v8(image detection), Mathpix OCR, Multilingual Embedding Model (LaBSE) and ElasticSearch resulted in 2X Daily Active User (DAU) for the product.
• Question Answering: Delivered New Book Digitisation Tools to the Embibe Content Team consists of OCR, Question Detection from Text, Answer Generation resulting in launching new books & courses 30X faster on Embibe platform.

Jan 2020 – Jun 2022 · 2 yrs 5 mos · Bengaluru Area, India

• Built model training and inference pipelines for construction progress tracking solution that uses computer vision to analyze images and videos from construction sites, comparing it against BIM models and schedules to provide insights such as quantities installed, percent completion, and identifying deviations from plans.
• Deep Learning: Trained Multi-View 3D Object Detection Model to generate embeddings of 3D Object Meshes using PyTorch on Google AI platform.
• Machine Learning: Worked on Feature Engineering (FPFH, IOU), Training & Evaluation of Random Forest model (Scikit-Learn) for 3D Object detection.
• Data Science: Worked on Spatial Data Science using PostGIS (spatial database extender for PostgreSQL), Numpy, Pandas.
• Data Engineering:
• Built backend APIs of the Data labelling platform using PostgreSQL, gRPC, REST API.
• Developed an annotation audit framework to build good Ground Truth datasets using PostgreSQL, SQLAlchemy, Pydantic, FastAPI, Alembic.
• ML Engineering:
• Integrated 3D Multiview Model with Weights & Biases Artifacts for Model and Dataset version control.
• Integrated 3D Multiview Model with W&Bs Sweeps to automate hyperparameter optimization.
• Integrated with Aquarium AI (Notion AI) to perform error analysis of 3D Multiview Model.
• Built ML training pipeline for 3D object detection using Apache Airflow DAGs, Docker and Kubernetes Engine on Google Cloud Platform (GCP).

Jan 2017 – Dec 2019 · 2 yrs 11 mos · Bengaluru Area, India

• Led team of 8 Engineers to develop HCI (Hyper Converged Infrastructure) capabilities for newer Hypervisor (Hyper-V 2019) and newer hardware platforms (with NVMe drives) that resulted in an annual revenue of USD 50M+.
• Collected CPU & Memory usage data of Cluster Nodes, VMs for capacity planning and anomaly detection using different time-series models like ARIMA, LSTM.
• Karbon Built Golang SDK of RESTful APIs using Swagger & CLI using Cobra for Nutanix Karbon, that deploys & manages Kubernetes cluster on Nutanix infrastructure.
• Xi Clusters: Worked on Hibernate & Resume of a running Nutanix Cluster on AWS bare metal using AWS APIs in Python.

Nov 2009 – Dec 2016 · 7 yrs 1 mo · Bengaluru Area, India

• Led data protection teams for Hyper-V 2016, SharePoint 2013 & SQL 2012 of EMC Networker using C++, C# & PowerShell.
• I have 20+ patents granted by USPTO for the innovations done in distributed backups, incremental backups, synthetic full restore, granular restore, instant start of big data applications.
• My blog on Data Protection Technologies for Microsoft Private Cloud: https://education.dellemc.com/content/dam/dell-emc/documents/en-us/2015KS_Chakraborty-Data_Protection_Technologies_for_Microsoft_Private_Cloud.pdf

Sep 2007 – Oct 2009 · 2 yrs 1 mo · Hyderabad Area, India

• Worked on implementing new functionalities for VPN & Routing (RRAS) in Windows 7 using C++, C# & PowerShell.
• My blog in technet: https://blogs.technet.microsoft.com/rrasblog/2009/07/02/how-to-configure-network-load-balancing-nlb-based-cluster-of-vpn-servers/

Feb 2006 – Aug 2007 · 1 yr 6 mos · Kolkata Area, India

• Led development team of 3 Engineers for developing NFT Corporation-AreaGen that generates Area Information from Layout Data for Proximity Error Correction (PEC) of Electronic Beam Machine using C++.
• Worked on Jaguar which is a high performance VHDL analyser using C++.

Apr 2003 – Dec 2003 · 8 mos · Kolkata Area, India

• Developed keyboard, mouse, display drivers for CEPC using Windows CE .NET & Platform Builder.
• Worked on SWIG to integrate code written in C/C++ with Python.

Jul 2002 – Mar 2003 · 8 mos · Kolkata Area, India

• • Implemented synchronization functions for concurrent healthcare claims processing and management on Lotus Notes & Domino Server.