Posted 7h ago
Retrieval Augmented Generation Prompt Intern
Posted yesterdayBe an early applicant
What this role pays, in context
This listing pays 15% below the median for AI & Prompt Engineering roles in India, which sits at ₹20.5K a month.
25th pct ₹19.3K75th pct ₹22K
Median of 5 other live AI & Prompt Engineering listings in India on this board that state pay. Roles that do not disclose a salary are excluded rather than counted as zero.
Ask our in product assistant about a job listing and it answers from a vector index of roughly 60,000 documents. The retrieval half is decent. The prompting half, honestly, is not.
The half you take over
Across three months you will own the generation side of our RAG pipeline: the system prompt that governs citation, the chunk formatting that reaches the model, and the fallback behaviour when retrieval returns nothing useful. A normal fortnight might start with reading 50 logged conversations where the assistant invented a company name, then tracing each one back to a bad chunk or a loose instruction. From there you rewrite the citation format, rerun the offline eval set, and ship the version that reduces unsupported claims. LangChain holds the pipeline together and you will spend real time inside its retriever and prompt template code.
Remote by default
This role is fully remote within India. We sync twice a week on video and everything else happens in writing, so clear notes matter more here than talking well in meetings. Your design documents are how decisions actually get made, and you will get feedback on them from week one.
Responsibilities
- Own the generation prompts for our document assistant, including citation and refusal behaviour
- Tune chunk formatting, ordering and metadata injection before text reaches the model
- Read logged conversations weekly and classify failure modes into a shared taxonomy
- Build an offline eval set of grounded question and answer pairs and rerun it before every prompt change
- Reduce unsupported claims by testing hedging and citation instructions against real traffic
- Document each prompt revision with its motivation and its measured effect
Requirements
- Final year student or recent graduate; no professional experience expected
- Working Python, enough to read and modify an existing LangChain pipeline
- Basic understanding of embeddings and vector similarity search
- Clear written communication, since most team decisions happen in documents
- Reliable internet and overlap with Indian working hours
Skills
Benefits
- Fixed monthly stipend with no attendance conditions attached
- Fully remote schedule with only two fixed sync calls a week
- Direct mentorship from the engineer who built our retrieval layer
- Certificate and a written reference on completion
- Your prompt changes ship to real users rather than a sandbox