Hi there, I'm Soumyadip

Data Scientist

MSc + BS Data Science graduate with 3+ years building end-to-end AI/ML solutions — LLMs, RAG, ML pipelines & data platforms — across telecom, pharma, manufacturing, consulting and sports.

View My Work arrow_forward Download Resume download

work_history 3+ Years Exp.

emoji_events 3X Kaggle Expert

groups 0+ Students Mentored

code 0+ LeetCode Solved

terminal Python

hub PyTorch

database SQL

monitoring Streamlit

Portfolio Narrative

About Me

I build end-to-end AI/ML systems that turn messy, real-world data into products people actually rely on — from LLM copilots to real-time analytics pipelines.

Over 3+ years I've shipped Retrieval-Augmented Generation tools, agentic LLM systems, forecasting engines and streaming data pipelines across telecom, pharma, manufacturing, consulting and sports domains — always tying the technical work back to measurable business impact.

I'm a 3X Kaggle Expert and mentor to 200+ students, and I care deeply about cross-functional collaboration, stakeholder communication, and building AI solutions that actually scale in production — not just in notebooks.

Education

BS in Data Science and Applications

Indian Institute of Technology, Madras · 2021–2025

MSc Data Science

Christ University, Bangalore · 2021–2023

BSc Statistics

University of Calcutta · 2018–2021

public 5 Industries Served

work GE Vernova Currently At

Work Experience

Chronicle / 2021–Present

01

Jun 2025 — Present

Data Scientist

GE Vernova · Bangalore

check_circle
Built a HyDE-based advanced RAG tool, cutting plant application onboarding from 6+ weeks to under 2 weeks for new configurators.
check_circle
Developed a LangGraph-based agent with LLM + SLM architecture, enabling talk-to-data across multiple real-time data sources.
check_circle
Achieved 97% text classification accuracy using SGD + Logistic Regression and LoRA-finetuned Llama3.2:1B, reducing LLM hallucinations.
check_circle
Created an end-to-end real-time Kafka → Druid pipeline, scaling to 300+ control loops, powering dashboards tracking 40+ live KPIs.

Python LLM / SLM RAG LangGraph Apache Druid Kafka Docker Vector DB

02

Jan 2024 — Jun 2025

Data Scientist

KPMG · Bangalore

check_circle
Enhanced the algorithm to parse and chunk PDFs, implemented RAG to deliver coherent responses, cutting processing time by 40%.
check_circle
Designed an automated time-series forecasting tool using GPT-3.5 and Prophet, automating 80% of the work.
check_circle
Integrated open-source LLMs like Llama-3 into the existing architecture, resulting in a 30% reduction in operating costs.
check_circle
Optimized inventory parameters using Monte Carlo Simulation, hitting target service levels and cutting cost by $500k–$750k.

Python SQL LLM / RAG Optimization FastAPI Azure ML Chroma

Earlier at KPMG — Data Scientist Trainee (Feb 2023 – May 2023): completed training in Python, SQL, ML, FastAPI, GenAI & Power Apps; achieved 92% accuracy in KPMG's internal ML hackathon; built a Power Apps prototype for leadership automating 5+ dashboard workflows.

03

Mar 2022 — Aug 2022

Data Science Intern

Tweek Labs · Bangalore

check_circle
Formulated an aggregated scoring algorithm ranking 500+ players by stats to help organizations build leaderboards.
check_circle
Streamlined the data storage pipeline via Google Sheets API and built an interactive dashboard with Meta Base tracking 15+ KPIs.
check_circle
Applied KNN and a stacked model to predict ground contact with average accuracy of 11ms for precise leg parameters.

Python Meta Base Google Sheets API Machine Learning

04

Sep 2021 — Oct 2021

Summer Intern

Ericsson · Kolkata

check_circle
Ran Root Cause Analysis, predicting root causes and recommending resolutions, reducing overall processing time by 50%.
check_circle
Wrote a generalized parser for XML log files, producing clean JSON for Elasticsearch using Python and Beautiful Soup.

Python Beautiful Soup VS Code

Projects

Personal build focused on applying LLMs to real academic workflows. Featured below — more projects incoming.

Oct 2024 — Dec 2024

Milestone.ai

AI-powered project tracker with role-based access control (RBAC) enabling instructors to track student progress and analyse project code & reports. Uses an LLM to generate structured milestones, checklists, descriptions and deadlines from a project-details PDF, fetches code via the GitHub API, and analyses it to produce personalized feedback.

Python FastAPI Vue Llama-3 LangChain JS SQLite RBAC GitHub API

code Source Code open_in_new App Link

Technical
Expertise

Languages & Frameworks

terminal

Python96%

SQL93%

FastAPI / Flask88%

LangChain / LangGraph85%

PyTorch / scikit-learn82%

PySpark90%

Machine Learning

psychology

95

Regression & Classification

88

NLP / Transformers

80

ANN / RNN / LoRA

90

EDA / Clustering

Tools & Technologies

database

Azure ML Tableau GCP A/B Testing Kafka Docker Apache Druid Ollama Streamlit Git RAG Hypothesis testing Chroma Time Series Forecasting Kubernetes

What I Offer

Services

Practical, production-minded data science — from raw pipelines to deployed intelligence.

smart_toy

LLM & RAG Systems

Designing retrieval-augmented generation pipelines, agentic workflows and fine-tuned SLMs that ship to production.

hub

ML Pipeline Engineering

Building reliable, scalable machine learning pipelines from raw data to deployed model.

timeline

Data Analysis & Forecasting

Time-series forecasting, statistical modelling and optimization to inform real business decisions.

school

Technical Mentoring

1:1 and cohort mentoring on data science fundamentals, interview prep and ML career paths — 200+ students mentored to date.

Book a Session arrow_outward

dashboard_customize

Dashboarding & Analytics

Interactive dashboards with Streamlit and Tableau that turn complex data into clear, actionable narratives.

Why Work With Me

What I Bring
to a Team

groups

Mentored 200+ students in data science fundamentals, interview prep and applied ML.

campaign

Led the ML team at GDSC, and was solo winner of two consecutive Hour-of-Code hackathons at KPMG.

podium

Featured speaker at the GE Vernova Users Conference 2026, presenting real-time AI systems in production.

format_quote

"Quote pending — replace with real feedback once collected."

Manager, GE Vernova (placeholder byline)

Gallery

Off Duty

Life outside the notebooks — trips, friends, and the odd stray moment. Click a photo to zoom in.

Data Scientist