Open to full-time roles & freelance

I build AI tools
that work in production.

|

Data Analyst & AI Developer turning messy problems into clean solutions. From data pipelines to deployed LLM apps — I build things that actually ship.

0+ Live Apps
0+ Projects
0 OSS Library
View My Work GitHub
Open Source Library

llm-behave

pytest plugin for behavioral testing of LLM applications

Stop using LLMs to judge LLMs. llm-behave brings semantic assertions, tone detection, and drift detection to your CI pipeline — all with an offline 80MB model. No API costs. No circular dependencies.

  • Semantic matching — response.mentions("refund") understands meaning, not just strings
  • Tone & intent detection — response.tone("professional") without an LLM judge
  • Drift detection for CI — catch behavioral regressions before they hit production
  • 100% offline — sentence-transformer runs on CPU, ~32ms per assertion
View on GitHub Read Docs ↗
test_support.py 
# Install the library
$ pip install llm-behave[semantic]

# Write semantic tests, not string matchers
def test_support_response(llm):
    response = llm.ask(
        "How do I get a refund?"
    )

    # Understands meaning, not just words
    assert response.mentions("refund policy")
    assert response.tone("professional")
    assert response.intent("helpful")

# Detect drift across model updates
def test_no_regression(drift):
    drift.assert_consistent(baseline="v1.json")
In Production

Live Applications

Real tools solving real problems. Try them.

LIVE
100+ API docs generated
DocGen Pro Interface
Try Live Demo →

DocGen Pro

Transform OpenAPI specs into beautiful documentation in seconds. 90% faster than writing docs by hand.

ReactNode.jsOpenAPI
Launch App ↗ View Code
LIVE
700+ datasets analyzed
Data Quality Detective
Try Live Demo →

Data Quality Detective

Upload a CSV, get instant AI-powered data quality scoring and actionable insights.

PythonStreamlitPandas
Launch App ↗ View Code
BETA
AI-Curated News
ClearFeed News
Try Beta →

ClearFeed

AI-curated news aggregation with smart categorization, real-time updates, and consensus ratings.

Next.jsAI/MLAPIs
Try Beta ↗ View Code
Work

Projects & Case Studies

Analysis, AI implementations, and technical deep-dives.

AI Bill Shock Detector 90% Accuracy

AI Bill Shock Detector

ML anomaly detection system for unexpected billing events. Real-time alerts preventing revenue loss.

PythonScikit-LearnStreamlit
View on GitHub ↗
Chat Assistant RAG Enterprise RAG

Chat Assistant – RAG

Private RAG implementation with PDF ingestion. 100% data privacy with instant knowledge retrieval.

LangChainOpenAIFAISS
View on GitHub ↗
InsightIQ 10x Faster

InsightIQ

Upload CSV/Excel, get GPT-4 powered business insights and automated reports instantly.

PythonGradioGPT-4
View on GitHub ↗
Multi-Agent

Accessible AI Agents

Multi-agent AI system for automated digital accessibility checks and content remediation.

PythonAgentsAutomation
View on GitHub ↗
Prompt Engineering Playground

Prompt Engineering Playground

Comprehensive demos of CoT, few-shot, RAG, and agent prompting techniques with benchmarks.

PythonOpenAIJupyter
View on GitHub ↗
Business Insights Chatbot

Business Insights Chatbot

Fully customizable AI chatbot using Hugging Face models — no OpenAI dependency required.

JavaScriptHuggingFaceAPI
View on GitHub ↗
Walmart Sales Analysis $2M Insights

Walmart Sales Analysis

Comprehensive sales data analysis identifying $2M in optimization opportunities with predictive modeling.

PythonPostgreSQLExcel
View on GitHub ↗
Financial Consumer Complaints 50K+ Records

Financial Consumer Complaints

EDA on 50K+ complaint records. Identified top categories driving regulatory action and process fixes.

PythonPandasPower BI
View on GitHub ↗
Student Data Analysis

Student Data Analysis

Power BI dashboard analyzing student performance, income distribution, and programming experience trends.

Power BIDAXExcel
View on GitHub ↗
Low Carbon Technology

Low Carbon Technology Trade

EDA of global low-carbon tech trade data (1994–2023) tracking sustainability trends and market shifts.

PythonMatplotlibPandas
View on GitHub ↗
Motorcycle Parts Sales

Motorcycle Parts Sales

SQL-driven analysis to optimize inventory, boost revenue, and improve customer retention strategies.

MySQLExcelSQL
View on GitHub ↗
Toolkit

Skills & Stack

Tools I use to get things done.

Databricks Generative AI Fundamentals Certified 2024

Languages

Python Python
SQL SQL
JavaScript JavaScript
NumPy NumPy

AI & LLMs

LC LangChain
OAI OpenAI
Scikit-Learn Scikit-Learn
HF HuggingFace
RAG RAG / FAISS
PE Prompt Eng.

Data & Analytics

Pandas Pandas
Matplotlib Matplotlib
PBI Power BI
TAB Tableau
Jupyter Jupyter
Spark Spark

Web & Infra

React React
Next.js Next.js
FastAPI FastAPI
Flask Flask
Docker Docker
SL Streamlit

Cloud & Databases

AWS AWS
Azure Azure
GCP GCP
MySQL MySQL
SF Snowflake
BQ BigQuery
Swanand Potnis
Available for projects
About

Hi, I'm Swanand.

I'm a Data Analyst and AI Developer based in India. My work sits at the intersection of data engineering and applied AI — I build tools that go from prototype to production, not just notebooks that sit on GitHub.

I started in financial analytics and moved into AI-powered solutions. Today I maintain llm-behave (an open source testing library for LLM apps), run 3 production applications with real users, and help businesses automate their data workflows.

Databricks Generative AI Certified (2024)
3 live applications with active users
Open source library author
Actively seeking full-time roles & open to freelance
Download Resume Let's Talk
Contact

Let's build something.

Actively looking for full-time roles in data analytics and AI. Also open to freelance projects and interesting collaborations — data analysis, AI solutions, process automation.

India — Remote worldwide
Actively seeking full-time roles
Also available for freelance projects