Hi, I'm Kowshik 👋

A passionate software engineer specializing in Machine Learning and AI. Currently working on building innovative AI solutions and exploring the intersection of technology and problem-solving.

I'm currently focused on developing AI applications and working with Large Language Models. My recent work includes projects in stock analysis, document processing, and generative AI applications.

Experience

AI/ML Intern - Fold (2586Labs)

Aug 2024 - Present

Implemented ML models for their F1 categorization engine; set up MLflow pipelines for better tracking of the implemented models.

Implemented an API endpoint for normalizing merchant patterns from transaction narrations with embeddings using Qdrant for vector-based similarity search.

Developed an advanced LangGraph-based agent with multi-tool integration, enabling precise financial data analysis and personalized insights through dynamic, context-aware query resolution.

Data Analysis Intern - Enrich

Oct 2023 - Dec 2023

Developed data models and utilized SQL on GCP (BigQuery). Conducted data visualization with LookerStudio, automating live data reports by integrating BigQuery with LookerStudio.

Generative AI Intern - VectrFlow

Aug 2023 - Sep 2023

Developed low-code platform for Generative AI apps, integrating Langchain and LlamaIndex frameworks. Specialized in fine-tuning LLMs for Image and Document Querying, as well as various generative AI models.

Projects

StockAI - Stock Analysis AI Agent

Developed an AI Agent that analyses stock prices, calculates probability using trading theories and provides latest finance news based on user input.

AI Python LangChain Finance

Advanced RAG AI Assistant

Developed an advanced RAG assistant using langchain and Streamlit, which ingests documents, stores them in Qdrant vector storage, indexes them, and performs user queries for precise, context, chat history-aware answers.

RAG LangChain Streamlit Qdrant

Data-Driven Energy Management

Analyzed weather-energy correlation in residential buildings to enhance energy management. Utilized models such as LSTM and BiLSTM to predict power consumption.

LSTM BiLSTM Python Data Analysis

Real-Time Chat Application

Implemented Real-Time Chat application in Python using Socket programming, enabling File transfer, multiple client support, and end-to-end data encryption.

Python Socket Programming Encryption