Forecasting Pipeline & React Integration
Summary
Freelancer Client is hiring: Forecasting Pipeline & React Integration.
Location: Remote
I have the Replenish.ai React + TanStack Start UI already in place; now I need it wired to a real-time forecasting backend that can service 100 stores and roughly 10 k SKUs per store from a daily POS feed.
What you'll do:
• Expose the pipeline through createServerFn so the existing React storefront can request forecasts in real time.
Skills: Python, Data Processing, Machine Learning (ML), Redis, AngularJS, JSON, API Development, Terraform
Budget: $250–$750 USD
Source: Freelancer Client via Remote / Online. Apply on the source website.
Original
I have the Replenish.ai React + TanStack Start UI already in place; now I need it wired to a real-time forecasting backend that can service 100 stores and roughly 10 k SKUs per store from a daily POS feed.
Scope
• Ingest the feed in CSV first, with the code structured so we can swap to JSON or Parquet later without touching business logic.
• Store raw and feature-engineered data in partitioned Parquet inside S3, queried locally through DuckDB/Polars and cached in Redis.
• Train and serve forecasts with LightGBM and Croston/TSB, orchestrated by Airflow.
• Ship a single-node pipeline that is production-ready yet cleanly abstracted for a 2–4 week lift to PySpark on EMR Serverless when volumes grow.
• Expose the pipeline through createServerFn so the existing React storefront can request forecasts in real time.
What must be fully exercised by end-to-end tests (Playwright + Vitest + pytest):
• Data ingestion & processing
• Machine-learning prediction paths
• Data storage & retrieval
• UI-to-API round-trips across the entire flow
Deliverables
1. Terraform definitions and IAM policies for S3, Redis/ElastiCache, Lambda endpoints and Airflow.
2. Python package (DuckDB/Polars, LightGBM, Croston) with unit tests and typed docs.
3. Airflow DAGs ready to deploy, parameterised for store/SKU segmentation.
4. TypeScript server adapters that plug straight into the TanStack Start frontend.
5. Playwright, Vitest and pytest suites running in CI, green from ingestion to on-screen forecast.
6. A concise migration guide outlining what changes when we switch the compute engine to Spark.
Please bid only if you have hands-on experience scaling a single-node ML pipeline to Spark and can share references of similar projects.
Location & Details
Apply on source →About this listing
This remote opportunity was imported from Freelancer and is shown here for discovery. To apply, follow the link to the original posting.