Skip to content
View sharangdev75's full-sized avatar
💻
Coding
💻
Coding

Block or report sharangdev75

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
sharangdev75/README.md

👋 Hey, I'm Sharang

I build stuff with data and AI. Been doing this for about 4 years now, mostly focused on turning messy data into things that actually help people make decisions.

What I Do

I'm a Data Scientist who's really into GenAI, machine learning, and making analytics systems that people actually want to use. Not just building models for the sake of it – I care about whether they solve real problems.

What I've Been Working On

GenAI & Agentic Systems (at OpenSky via CeADAR)

Built a Text-to-SQL system that doesn't completely break when someone asks a weird question. Used Azure OpenAI with some guardrails so it wouldn't generate nonsense queries. Cut SQL errors by 40%, which was pretty cool.

Made a regional similarity recommendation engine using kNN across 130+ regions. Analysts got their answers 35% faster, which means fewer angry Slack messages.

Put together a RAG platform for internal docs. No more "hey where's that document?" in every meeting. Saved about 30% of the time people spent hunting for information.

Built Power BI dashboards with natural language queries for a pharma client. They can now ask questions in plain English instead of learning DAX. Reporting got 30% more efficient.

Machine Learning & BI (at NextraIT)

Built customer retention models with Random Forest and XGBoost. Got a 15% boost in retention, which made the business folks happy.

Created executive dashboards that actually got used (shocking, I know). Decision-making got 20% faster.

Did the usual ML stuff – classification, anomaly detection, threshold optimization. Made sure to pick the right metrics instead of just chasing accuracy.

Cloud Analytics & Forecasting (at Invoice Fair)

Improved forecasting accuracy by 12% using ARIMA and SARIMAX. Played around with LSTM too, just to see.

Built 15+ Tableau dashboards for sales and customer success teams. Real-time updates, so they're not looking at yesterday's data.

Set up AWS pipelines with Glue, Athena, S3, and QuickSight. Data became 30% more accessible and we cut cloud costs by 10%. Win-win.

Currently Learning

Right now I'm going deeper on AWS analytics (Glue, Athena, QuickSight, DynamoDB), Microsoft Fabric, and agentic AI systems. There's always something new to figure out.

Things I Know Pretty Well

GenAI Stuff: Text-to-SQL, RAG systems, prompt engineering, making sure LLMs don't go off the rails

Machine Learning: Time series forecasting, classification, anomaly detection, actually understanding what the metrics mean

BI & Dashboards: Power BI (including DAX and that AI Q&A feature), Tableau, designing KPIs that make sense

Cloud Platforms: AWS and Azure data tools, building pipelines that don't fall over

Finance & Risk: Fraud detection, credit risk modeling, portfolio stuff

Tech Stack

Languages: Python, R, SQL

ML & Data: Pandas, NumPy, Scikit-learn, NLP libraries, statistical modeling

BI: Power BI, Tableau, Excel (yes, it still counts)

Cloud: AWS (Glue, Athena, S3, QuickSight), Azure (SQL, OpenAI, Synapse), Microsoft Fabric

GenAI: Azure OpenAI, LangChain, building RAG systems, API integrations

Let's Connect

Want to talk about GenAI, ML projects, or just geek out about data? Hit me up:

📧 Email: [email protected]

💼 LinkedIn: linkedin.com/in/sharang75


P.S. If you're looking at my repos and wondering why some are messy – they're experiments. That's what GitHub is for, right?

Popular repositories Loading

  1. RFM-Values-using-K-Means RFM-Values-using-K-Means Public

    Calculating RFM values

    Jupyter Notebook 3

  2. HR_Domain HR_Domain Public

    Jupyter Notebook 2

  3. ANALYSING-THE-CRIMES-IN-INDIA ANALYSING-THE-CRIMES-IN-INDIA Public

    In the article in the Business Insider, suggests that the crimes in India have cost economic growth. According to an international think tank, the Institute for Economics and Peace, crimes have cos…

    Jupyter Notebook 2

  4. SA_Heart_Disease SA_Heart_Disease Public

    Jupyter Notebook 1

  5. Prediction-on-Religion- Prediction-on-Religion- Public

    R 1

  6. Multiple-Linear-Reg Multiple-Linear-Reg Public

    R 1