Machine Learning Intern

Role Overview

We’re looking for a Machine Learning Intern who’s excited about building models that solve real Trust & Safety problems at scale. As part of the FRND ML team, you’ll train and ship models that help identify fake profiles, unsafe interactions, abusive speech, harassment, spam, and scams on the platform. You’ll work with production data alongside experienced engineers, and the models you build will run on live traffic, not just in a notebook.

Responsibilities

• Work on image & video ML - fake-profile detection, and spoofed-camera detection on 1:1 video calls.
• Work on audio ML - abuse and unsafe-speech classification, multilingual and code-mixed ASR for audio rooms.
• Work on text ML - harassment, spam, and scam detection across chat in Hindi, English, and regional languages.
• Build and clean datasets from production signals and moderation reports, and help define labelling guidelines.
• Train and fine-tune models in Python with PyTorch - including vision backbones, audio encoders, and transformer-based text models.
• Evaluate models with a production mindset - precision at fixed recall, per-language performance, and the real cost of false positives.
• Collaborate with backend engineers to turn models into inference services and monitor latency and throughput in production.
• Monitor model drift, review misclassifications with the moderation team, and improve models through better data and retraining.
• Explore LLMs for labelling, policy classifiers, and model evaluation.
• Write clean, reproducible code and maintain well-documented experiments.

Requirements

• Available for a 6-month, in-office internship.
• Strong academic background from a Tier 1 institution, preferably IITs, BITS, NITs, or equivalent institutions.
• Pursuing a degree in Computer Science, Engineering, Mathematics, Statistics, or a related field.
• Strong Python skills and solid programming fundamentals.
• Working knowledge of PyTorch or TensorFlow.
• Strong understanding of ML fundamentals - overfitting, regularisation, train/validation/test hygiene, precision vs. recall.
• Comfortable with NumPy, pandas, and notebooks for data work.
• Projects in computer vision - CNNs, ViTs, classification/detection - are a plus.
• Exposure to audio ML - spectrograms, ASR, Whisper, or wav2vec-style models - is a plus.
• Experience with NLP, transformers, or Hugging Face is a plus.
• Basics of SQL, Git, or Docker are a plus.

Skills

• Python
• PyTorch
• TensorFlow
• NumPy
• pandas