AI/ML Engineer — i3w Careers

About the Role

i3w ships real ML systems, not thin API wrappers. We build predictive models for government child nutrition programs (Poshan-AI), safety-filtered AI tutors for students preparing for NEET and JEE (Companion), and RAG pipelines that generate curriculum-aligned study material. Our stack runs on GCP with Vertex AI, Gemini models, and Cloud SQL for structured data. This is applied ML with production constraints, not research for the sake of benchmarks.

As an AI/ML Engineer, you will work directly under Bharat Agarwal (ex-Atlassian/Shopify) on two verticals: Companion (a voice-first AI tutor with dual WebRTC + WebSocket architecture) and B2G systems (Poshan-AI malnutrition prediction, Angan AI curriculum generation). You will implement models, build safety filters for student-facing AI, and own ML pipelines from prototype to production. Evaluation frameworks are a shared responsibility — you will work closely with Rishabh on evals, with Bharat setting the direction.

We use Claude Code as our primary development tool, and we expect you to be fluent in AI-augmented development. But this role is specifically about the ML layer — you are building the intelligence, not just consuming it. If you want to work on real prediction problems with real data and real deployment stakes (government systems serving actual children), this is the role.

What You'll Do

Build and maintain RAG pipelines for curriculum-aligned content generation (NCERT/CBSE syllabus, regional language support)
Implement and fine-tune safety filters and guardrails for student-facing AI (content appropriateness, factual accuracy, age-appropriate responses)
Develop evaluation frameworks: automated evals, human-in-the-loop evaluation, regression testing for model outputs
Build predictive ML models for Poshan-AI — risk scoring for child malnutrition using noisy government health data
Design and implement the voice AI pipeline for Companion: speech-to-text, LLM reasoning, text-to-speech, with Hinglish support
Work with Gemini models via Vertex AI, including prompt engineering, context window management, and output parsing
Set up ML monitoring: model drift detection, output quality tracking, latency and cost optimization
Collaborate with the frontend and backend teams to integrate ML services into production applications via Cloud Run APIs

What We're Looking For

1-3 years of experience in applied ML or NLP (industry or strong research experience)
Solid Python fundamentals and experience with ML frameworks (PyTorch, TensorFlow, or scikit-learn)
Hands-on experience with LLMs: prompt engineering, RAG architectures, embedding models, vector stores
Understanding of evaluation methodology — you can design an eval suite, not just run one
Comfort with cloud ML infrastructure (GCP Vertex AI preferred; SageMaker or Azure ML also fine)
Ability to read papers and implement what matters, skip what doesn't — practical ML judgment
Strong communication: you will explain model behavior and tradeoffs to non-ML teammates

Nice to Have

Experience with speech/voice AI: ASR (Whisper, Deepgram), TTS, WebRTC audio pipelines
Exposure to Gemini, Claude, or GPT-4 APIs in production (not just prototypes)
Experience building safety/moderation systems for consumer-facing AI
Familiarity with Indian language NLP: Hindi, Odia, Hinglish code-switching
Background in EdTech, health-tech, or government/public-sector ML
Experience with ML experiment tracking (Weights & Biases, MLflow) and model versioning
Published work or open-source contributions in NLP/ML

Apply: bharat@i3w.ai