Website:
khwaaish.com
Job details:
Prompt & LLM Engineer Intern — Khwaaish
Prompt & LLM Engineer Intern
We are hiring a Prompt & LLM Engineering Intern with 1+ year of hands-on experience to work on improving AI accuracy, query understanding, and product relevance across our quick-commerce platform. The role requires strong intuition for language models, prompt design, and the ability to diagnose and fix real-world LLM failures in production.
Internship Details
• Duration: 3 months
• Work mode: Fully remote
• Opportunity: Unpaid currently, High-performing interns may be considered for full-time roles after the internship
About Khwaaish
Khwaaish is a technology platform that aggregates multiple quick-commerce services. We use LLMs at the core of our product to understand user queries, extract intent, and rank the most relevant products in real time.
Experience Required
• Minimum 1+ year of hands-on experience working with LLMs or prompt engineering
• Experience with real-world AI systems, not just academic or tutorial projects
Key Responsibilities
• Design, test, and iterate on prompts for query understanding and product relevance
• Analyse LLM failures from production logs and write targeted fixes
• Build and maintain few-shot example libraries for Indian grocery and quick-commerce queries
• Improve extraction accuracy for product names, quantities, sizes, and brand names
• Work on Hindi/Hinglish query understanding and multilingual prompt robustness
• Evaluate model outputs systematically and track accuracy improvements over time
• Collaborate with backend engineers to integrate prompt changes into the live pipeline
• Design fallback logic when LLM confidence is low
• Maintain prompt versioning and document changes with before/after accuracy metrics
Mandatory Skills
• Strong understanding of how LLMs work — context windows, temperature, few-shot prompting
• Hands-on experience writing and iterating prompts for structured output (JSON extraction)
• Ability to diagnose why a model is giving wrong outputs and fix it systematically
• Familiarity with at least one LLM API — Gemini, OpenAI, Anthropic, or similar
• Good written communication — prompts are essentially precise technical writing
• Experience with Python for scripting, testing, and log analysis
Good to Have
• Exposure to Indian language queries or vernacular NLP
• Understanding of retrieval-augmented generation (RAG) or semantic search
• Familiarity with evaluation frameworks for LLM outputs
• Knowledge of quick-commerce or e-commerce product data
• Experience with LLM observability and logging tools
• Understanding of fallback systems and confidence thresholding
How to Apply
Send your CV along with a brief note on an LLM failure you diagnosed and fixed — or a prompt you wrote that significantly improved model output — to tech@khwaaish.com
Click on Apply to know more.