Bhargob Deka, Ph.D.

Experience

• 

  NORSAR

  Jun 2014 - Jul 2015

  Research Engineer

  1. Worked in the EQRisk Project where we developed a seismic risk model of Guwahati city.2. Created 2-D and 3-D prototype models of different building typologies in SAP2000.3. Carried out pushover analysis to obtain load versus displacement curve for each model.

• 

  McGill University

  Apr 2016 - Jun 2018

  Research Assistant

  Statistical downscaling of weather variables from low resolution map to point stations using multivariate statistical analysis for climate change scenarios.

• 

  Hydro Québec

  Sept 2018 - May 2024

  Research Affiliate: Probabilistic Modeling and Anomaly Detection in Time Series

  1. Devised a multiplicative state-space model to handle non-harmonic periodic patterns often encountered in structural health monitoring applications.2. Developed an interpretable ML approach for modeling nonlinear dependencies affecting a concrete dam's displacement.3. Collaborated with a team of dam engineers from Hydro-Quebec to implement the ML model for analyzing real data from dams in Quebec, Canada.4. Combined the method with a real-time anomaly detection platform to trigger an alarm at the onset of any structural behavior change.5. Participated in the predictive modeling competition at the ICOLD-BW2022. Our ML model was ranked first and fourth for the two benchmark datasets, respectively, among all competitors. Show less

• 

  Polytechnique Montréal

  Jan 2023 - May 2024

  Machine Learning Researcher: Uncertainty Quantification for Machine Learning Models

  1. Currently, I am building a scheduler that would self-tune the learning process in a Bayesian neural network. It leverages AGVI and local acceleration model to help the model adapt its learning over epochs. 2. The model has been tested successfully for small-scale regression and time series datasets.When ready, it will seamlessly be applied to larger regression and classification architectures as well as for analyzing large number of time series by leveraging an LSTM network with the AGVI model.3. This model, if successful, would be able to bypass hyper-parameter tuning of error variance, one of the most computationally demanding task in model training. Show less

• 

  Medium

  Jun 2024 - now

  Technical Writer

  Helping fellow developers/student/businesses to learn Generative AI with hands-on practical articles and tutorials. Within 2 months, I have written a total of 𝟗 𝐚𝐫𝐭𝐢𝐜𝐥𝐞𝐬 + 𝐜𝐨𝐝𝐞 while garnering 𝟐𝟓𝟎+ followers and 𝟐𝟎+ subscribers on Medium.Currently, focusing on multi-modal RAG chat application and leveraging small language models/fine-tuned LLMs to help businesses.Feel free to contact me if you have any question regarding my articles/ codes or want any help in your buisnesses.Here is my profile: https://medium.com/@bhargobdeka11Created a beginner series of a total of 7 articles starting from the basics of Langchain: a framework that lets you build LLM apps, to deploying an end-to-end chatbot on AWS. 𝗧𝗵𝗶𝘀 𝗶𝘀 𝗶𝗻𝘁𝗲𝗻𝗱𝗲𝗱 𝗳𝗼𝗿 𝗮𝗻𝘆𝗼𝗻𝗲 𝘄𝗵𝗼 𝘄𝗮𝗻𝘁𝘀 𝘁𝗼 𝗴𝗲𝘁 𝗶𝗻𝘁𝗼 𝗔𝗜 𝗯𝘂𝘁 𝗱𝗼𝗻'𝘁 𝘄𝗵𝗲𝗿𝗲 𝘁𝗼 𝘀𝘁𝗮𝗿𝘁.From basic concepts to advanced deployment strategies, this series covers everything you need to start building your own AI applications using LangChain.Here’s a quick recap of what I’ve covered:1️⃣ Chatbots with LangChain: A gentle introduction with Python2️⃣ Creating Retrieval Chain: RAG chatbot with LangChain3️⃣ Creating RAG Agents with LangChain4️⃣ Designing RAG Agent Workflow with LangGraph5️⃣ LangGraph, FastAPI, and Streamlit/Gradio: The Perfect Trio for AI Development6️⃣ From Local to Cloud: Deploying LLM Application with Docker and AWS EC27️⃣ Cloud Deployment with a Git Push: GitHub Actions is All You NeedI have also posted articles on advanced AI topics such as creating multi-agent framework with two end-to-end projects. They are:8️⃣ Building a Hotel Recommender: Multi-Agent Framework with CrewAI, Ollama, and Gradio9️⃣ Creating an End-to-End Data Science Project with CrewAI Agents Show less

• 

  Jesta I.S.

  Oct 2024 - now

  Senior AI/ML Developer

  In this position I am working as a lead on Jesta's AI projects. The work includes building end-to-end AI Agent workflows and implementing ML projects at scale targeted towards the retail/fashion sector.