- by RohitschauhanitbhuWe all know Langchain from its start in October 2022 by Harrison Chase. For the uninitiated, Langchain is the open-source Python (and JavaScript) framework designed to make it easier to build applications powered by LLMs. It’s not just about calling an LLM like GPT-4 — LangChain helps you connect LLMs with external data, tools, and workflows, so they become useful in real apps. I will go into the workings and internals of langchain during this series of Medium articles focussing on the Langchain world, which at this point consists primarily of the Langchain, Langgraph and Langsmith.Before we start, for someone just starting […]
- by RohitschauhanitbhuStochastics Processes — Part 2: Chapman-Kolmogorov Equation and different states of a Markov chain, with applications in RIn part 1 of the Stochastics blog series (given below), we discussed about the mathematical introduction to a stochastic process, when a stochastic process becomes Markovian, how to understand the one-step transition probability matrix in a Markov chain along with calculating conditional and joint probabilities on it, and the plotting of a Markovian process in R using the diagram package.Stochastics Processes — Part 1 : Getting started with a mathematical introduction and simple…In this blog, we will take things one step further and move to n-step transition probability matrix, the […]
- by RohitschauhanitbhuStochastics Processes — Part 1 : Getting started with a mathematical introduction and simple plotting of Markov Chain in RA few months back, one of our clients from a big pharma company came to us and asked us to predict the enrolment rate for patients getting onboarded onto clinical trials for different drugs in development. The problem was simple: our client wanted to understand which patient enrolment sites were performing well/which needed to be shut down as different sites(or clinics) had different enrolment rates of patients getting enrolled for different drugs. Whatever was the profile of the drug, or the geography of the site, […]
