Data Science Course with Placement Support

Data Science Projects you’ll actually build

Python, SQL, Machine Learning, NLP, Retrieval-Augmented Generation, Agentic AI, and PyTorch serving—taught with small, practical briefs that read well and run locally.

1. Retail Analytics with Python & SQL

   Beginner • 1 week

   Clean sales data, run cohort and AOV analysis, and publish weekly KPIs using a small, reproducible Python + SQL pipeline.

   What you’ll practice

   • Data quality checks (missing values, outliers)
   • Reusable SQL queries and a simple ETL script
   • Clear notebook/report that a non-technical lead can read

2. Customer Churn: Model to Minimal API

   Intermediate • 2 weeks

   Train a baseline churn model and expose a minimal prediction endpoint. Focus on feature logic, evaluation, and a clean interface.

   What you’ll practice

   • Feature pipeline + ROC-AUC / PR-AUC evaluation
   • Small FastAPI service with schema-checked requests
   • Straightforward logging and a short “how to run” note

3. NLP Text Classification with a Small Transformer

   Intermediate • 2 weeks

   Fine-tune a compact transformer to categorise product reviews. Keep the preprocessing honest and the evaluation easy to interpret.

   What you’ll practice

   • Tokenisation + class balancing where needed
   • Confusion matrix + error examples for stakeholders
   • Simple batch inference script

4. RAG over Private Documents

   Intermediate • 2 weeks

   Search internal PDFs and answer questions with sources. Emphasis on chunking, retrieval quality, and grounded responses.

   What you’ll practice

   • Ingestion + embeddings + retriever with a vector index
   • Prompting that cites sources and avoids hallucinations
   • A tiny `/ask` endpoint; latency notes and trade-offs

5. Agentic AI: Plan → Research → Draft

   Advanced • 3 weeks

   Orchestrate a planner, researcher and writer. Add a verification step so drafts include links that can be checked quickly.

   What you’ll practice

   • Clear roles and a simple graph of steps
   • Tool use for search and citation capture
   • A “critique” pass that rejects weak sources

6. PyTorch to TorchScript, then Docker

   Advanced • 3 weeks

   Export a trained PyTorch model to TorchScript and serve it. Keep the container small, the health checks obvious, and the API stable.

   What you’ll practice

   • TorchScript export (script/trace) with parity checks
   • FastAPI server with health, version and simple batching
   • Dockerfile plus a short load test and latency note

Ready to join the next cohort? Register here