Supratik Vishnu — Business Analyst · Software Engineering

cat about.md

Summary

Computer Science graduate (PES University, 2022–2026) working across three overlapping areas: business analytics — AI-powered product experimentation, cohort analysis, and data visualization at scale on a ~65M DAU platform; software engineering — production SQL/Python pipelines, plus hands-on project work in socket programming, network protocols, and virtualized systems infrastructure; and machine learning — forecasting models, anomaly detection, and LLM-orchestrated systems.

Co-author of a peer-reviewed IEEE ISCAIE 2026 paper on an AI-orchestrated financial analytics platform, where all three of those areas come together — the work that anchors the featured research below.

grep -r "published" ./research

Featured Research

IEEE ISCAIE 2026 · Published

Cash Flow Forecasting and Analysis Leveraging LLM and Machine Learning Techniques

V. V. Ullasa, R. Rajendran, Supratik Vishnu, R. R. Rao, Dr. Nagegowda K S, R. A. Bangalore — Dept. of CSE, PES University

An intelligent cash-flow forecasting system for MSMEs combining a LightGBM recursive forecasting engine with a novel Per-Supplier Normalization technique, an Isolation Forest anomaly detector explained via LIME, a dual-level profitability framework (Prophet at macro level, LightGBM at micro level), and a ReAct LLM orchestrator that routes natural-language queries to the right analytical backend. Evaluated on three years of real retail transactions (2017–2019).

0.89

R² · cash-inflow

0.87

R² · cash-outflow

95.5%

query-routing accuracy

61%

↓ inter-vendor MAE bias

My contribution: the ReAct orchestration layer (LangGraph + Gemini 2.5 Flash) that parses natural-language queries and routes them to the correct ML backend via temporal and granularity-based validation, and the Flask API serving non-blocking requests across the LightGBM, Isolation Forest, and Prophet models.

tail -f experience.log

Experience

Business Analyst I

Jun 2026 – Aug 2026

Meesho · Bengaluru, India

Wrote SQL and Python to process experiment and behavioral data at scale on a ~65M DAU platform, building automated pipelines feeding GenAI feature experiments for Vaani, Meesho's AI shopping assistant.
Spearheaded prompt and feature experiments rolling out targeted GenAI features to retained users, and presented experiment results and adoption metrics in weekly leadership connects.
Built and maintained a metric-store computation pipeline (VRS) benchmarking pricing data across matched products — data cleaning, flag computation, and aggregation logic at scale.
Conducted post-launch analysis of GenAI experiments — conversation logs, search reformulation patterns, session behavior — to measure response quality, friction, and retention impact.

Business Analyst Intern

Jan 2026 – Jun 2026

Meesho · Bengaluru, India

Ran SQL-based cohort analysis on Non-GST seller segments to identify churn-risk segments and evaluate interventions, and built automated Superset dashboards tracking onboarding conversion, field sales performance, and regional KPIs.
Contributed to Explore, a feed pod surfacing fresh catalog content, by identifying high-performing catalogs with minimal view investment to improve feed diversity.

ps aux | grep projects

Projects

Cash Flow Forecasting & Analysis Platform

IEEE ISCAIE 2026

AI-orchestrated financial analytics platform built on 300K+ real-world retail transaction records. Designed the ReAct agent orchestration layer (LangGraph + Gemini 2.5 Flash) that parses natural-language queries and routes them to the correct backend tool using temporal and granularity-based logic, reaching 95.5% tool-routing accuracy; built the Flask API backend supporting non-blocking queries across LightGBM forecasting, Isolation Forest anomaly detection, and Prophet models.

LangGraph · Gemini 2.5 Flash · Flask · LightGBM · Isolation Forest · Prophet

TCP/UDP Chatroom

networking

TCP and UDP chatroom applications in Python enabling real-time multi-client communication over raw sockets; used Wireshark to compare latency, reliability, and packet-loss tradeoffs between the two protocols.

Python · Sockets · Wireshark

Proxmox Home Cloud Lab

infra

A home cloud built on Proxmox (Type-1 hypervisor) across two repurposed laptops as cluster nodes; deployed and orchestrated microservices with load balancing, automated backups, and snapshotting.

Proxmox · Microservices · Docker · Linux

./skills --list --verbose

Skills

Analytics & Strategy

Business AnalyticsCohort Analysis A/B TestingKPI Tracking ForecastingSupply Chain Analytics

Machine Learning

XGBoostRandom Forest LSTMProphet Scikit-LearnFeature Engineering

Engineering & Systems

PythonC++C GoJavaScriptSQL Socket Programming (TCP/UDP)Linux ConcurrencyFlask REST APIsDocker ProxmoxMicroservices

Visualization & Cloud

Power BITableau AzureMicrosoft Fabric SQL ServerMySQL

cat education.log

Education

PES University

B.Tech, Computer Science & Engineering · 2022 – 2026 · CGPA 8.26/10

Data Structures & Algorithms, Operating Systems, Computer Networks, DBMS, Object-Oriented Programming

CMR National Public School

Grade 12 · 2022 · 96%

CMR National Public School

Grade 10 · 2020 · 91%

Supratik Vishnu receiving his B.Tech degree certificate on stage at PES University convocation

// convocation — PES University, 26 July 2026

cat certifications.txt

Certifications

→ Linux Fundamentals — The Linux Foundation
→ Full Stack Machine Learning — MSME, Govt. of India
→ Python for Everybody Specialization (5 courses) — University of Michigan
→ PwC Launchpad Program

curl -I contact

Let's talk.

Open to software engineering roles in systems, backend, and infrastructure. Reach out directly — I read every message.

supratik.vishnu@gmail.com linkedin.com/in/supratik-vishnu github.com/SupratikV