Posted 1h ago
Loading this job…
Feature Engineering Intern (pandas + Python)
Posted 2h ago0 views
Features are the product here
Ask anyone on our data team where the last real accuracy gain came from and they will name a feature, not a model. This three-month internship puts you in charge of the feature layer that every MailerMen model reads from.
Concretely
You work in pandas and Python against roughly two years of posting, application and recruiter-response data. Expect time-windowed aggregates such as how many applications an employer answered in the trailing thirty days, text-derived signals pulled from job descriptions, and geography features that cope with the fact that one metro label can mean five different commutes. You also write the tests that stop a silently null column reaching training, which sounds dull right up until the first time it saves a release.
The setup
Fully remote within India and asynchronous by default, with two pairing sessions a week and review on every pull request. You need to be comfortable being the person who keeps asking where a column came from. Written for final-year students and recent graduates who like data more than architecture diagrams and have used pandas on something larger than a tutorial dataset.
Responsibilities
- Build time-windowed aggregate features such as employer response rate over a trailing thirty days
- Derive text-based signals from job descriptions for use across ranking and matching models
- Normalise geography features so metro regions covering several commute zones behave sensibly
- Write data tests that fail the pipeline when a feature column arrives silently null
- Maintain a feature dictionary recording the definition, owner and refresh cadence of each column
- Backfill new features across two years of history and validate the output against manual spot checks
- Pair with model owners to retire features that nothing reads any more
Requirements
- Final-year student or recent graduate happy to spend most of the day inside pandas
- Python, plus SQL for pulling and validating data straight from the warehouse
- Understands leakage and why a feature computed after the label is worthless
- Instinctively asks where a column came from before trusting it
- Available for two pairing sessions a week during Indian business hours
Skills
Benefits
- Monthly stipend with fully remote, asynchronous working inside India
- Twice-weekly pairing with a senior data engineer and review on every pull request
- The feature dictionary you build becomes documentation the whole team relies on
- Flexible hours around university deadlines and exam weeks
- Certificate of completion and a personal reference from your mentor