Posted 1h ago
Loading this job…
Applied Machine Learning Intern (Python + scikit-learn)
Posted 2h ago0 views
The problem you will work on
MailerMen ranks around 40,000 startup job posts against candidate profiles, and the ranking model we ship today still leans on hand-tuned weights. Over this six-month internship you will help replace those weights with a gradient boosted model trained on click and application data from the live board.
Month by month
Your first fortnight goes into pandas: pulling the raw event tables, spotting duplicate applications, and building a training frame you actually trust. From there you move into scikit-learn, cross validating candidate models against the current heuristic baseline and working out whether a lift in AUC survives contact with real traffic. Later on you own one full feature family, most likely the recruiter-response signals, from first idea to shipped production column.
Who this suits
Final-year students and fresh graduates who have finished a couple of competition-style projects and now want to see what changes when the data is messy and the users are real. You will sit two days a week with the search team in our Koramangala office and arrange the rest of your week yourself. By the end you will have a model in production, an error analysis nobody else has done, and a clear opinion about what ranking quality actually means on a job board.
Responsibilities
- Assemble a clean training dataset from the job board's click, save and application event tables using pandas
- Train and cross validate gradient boosted ranking models in scikit-learn against the current hand-tuned heuristic
- Own one feature family end to end, from hypothesis through offline validation to a shipped production column
- Run error analysis on the postings the ranker gets most wrong and bring findings to the weekly search review
- Record every experiment, including the ones that did not beat baseline, in the team's shared results log
- Watch the metric dashboards during A/B rollout for the first week after each release
Requirements
- Final-year student or recent graduate in computer science, statistics, mathematics or a related field
- Comfortable writing Python and using pandas for joins, groupbys and time-based aggregation
- Understands train and test splits, overfitting, and why a validation set exists at all
- Has at least one personal or academic project involving a trained model, with code you can walk us through
- Able to be in the Bengaluru office two days a week
Skills
Benefits
- Monthly stipend paid on the first working day, with no invoicing chase
- One-to-one mentorship from the engineer who owns the search ranking stack
- Full-time offer considered at the six-month mark based on what you shipped
- Internship certificate and a reference letter written by your mentor
- Laptop provided and travel reimbursed for office days