Generative AI, agents, LLM applications, computer vision and applied ML
The fastest-moving hiring category in India right now. Teams that shipped their first LLM feature in 2025 are staffing 2026 with interns who can write evals, build retrieval pipelines and ship agent workflows — not just train notebooks. Supply is thin at the senior end and unusually strong at the intern end, which is exactly why internships are the efficient way in.
Why this matters for 2026 hiring
AI Specialist and Generative AI Engineer are among the fastest-growing career-starting titles in India. Posting under a precise AI title — “Generative AI Intern”, “AI Agent Intern”, “RAG Intern” — reaches candidates that a generic “Data Science Intern” post never surfaces.
Each page carries a copy-ready JD, the stipend band across 64 cities, the skills worth screening for, and three questions that separate strong candidates from strong CVs.
A generalist AI intern who builds and evaluates AI-powered features — wiring models into a product, measuring whether the output is actually good, and improving it.
Builds product features on top of generative models — text, image or code — and owns the prompt, retrieval and evaluation layers that make them reliable enough to ship.
Builds multi-step AI agents that use tools, call APIs and complete tasks end to end — and the scaffolding that keeps them from going off the rails.
Works directly with large language models — evaluating them, adapting them to a domain, and squeezing quality, cost and latency into an acceptable range.
Owns the retrieval half of AI features — chunking, embedding, indexing and ranking company content so models answer from facts instead of memory.
A product engineer for AI features — builds the full surface around a model: the UI, the streaming, the state, the error handling and the feedback loop.
Turns fuzzy product requirements into prompts that behave consistently — and proves it with evaluation data rather than vibes.
Finds repetitive internal work and removes it with AI-assisted automations — the intern whose output is other people’s freed-up hours.
Trains, validates and deploys models on the company’s own data — classification, regression, ranking, recommendation and forecasting.
Works on neural network architectures and training pipelines where classical ML is not enough — vision, sequence, audio or multimodal problems.
Builds systems that read text at scale — classification, extraction, summarisation, sentiment and search over the company’s unstructured content.
Builds systems that interpret images and video — detection, classification, OCR, tracking and quality inspection.
Investigates open questions rather than shipping features — literature review, hypothesis, experiment, and a written result the team can act on.
The bridge between what AI can do and what users need — scopes AI features, defines quality bars, and decides what is worth building.
The Python-first entry point into AI work — writes the data, training and serving code that every other AI role depends on.
Applies AI to machines that move — perception, control, navigation and the simulation-to-reality gap.
Answers business questions with data — designs the analysis, builds the model where one helps, and tells the story that changes a decision.
A blended role for teams that need both classical analysis and AI feature work — analytics on weekdays, model integration when the roadmap calls for it.
Keeps AI deployments defensible — documents what models do, tests them for harmful or biased behaviour, and maps them against India’s emerging AI and data rules.
Builds the rails models run on — training pipelines, deployment, versioning and monitoring — so data scientists stop shipping by hand.
The same role costs materially less outside the metros for candidates of comparable quality. If the work can be done remotely, these markets are worth posting into.
Showing the first 10 of 20 roles. Every role page carries the full 64-city benchmark.
Reach students and freshers across 64 Indian cities. No posting fee, no limit on the number of roles.
Every listing reviewed before publishing