Posted 1h ago
Loading this job…
AI Agent Tooling and Prompt Engineering Intern
Posted 2h ago0 views
Agents fail in boring ways. They call the wrong tool, they call the right tool with a malformed argument, or they loop until the token budget is gone.
Where you come in
Over this six month internship you will work on the recruiter agent that screens applications, schedules calls and drafts follow up email. Your territory is the layer between the model and its tools: the tool descriptions, the argument schemas, the repair prompts and the stop conditions. You will rewrite tool docstrings until the model stops guessing, build a trace viewer so any failed run can be replayed step by step, and run structured experiments on how much scratchpad reasoning actually improves tool selection.
Reading traces
Expect to read a lot of them. Most of the insight in agent work comes from sitting with a failed run until the cause becomes obvious, and that habit is the main thing this internship teaches.
Support around you
You will pair with a senior engineer weekly and present one finding to the wider team each month. Nobody expects you to arrive knowing agent frameworks. We expect you to be stubborn about understanding why a run failed. The Pune team works from the office two or three days a week.
Responsibilities
- Write and refine tool descriptions and argument schemas for our recruiter agent
- Design retry and repair prompts that recover from malformed tool calls
- Build and maintain a trace viewer that replays agent runs step by step
- Run controlled experiments on reasoning depth versus tool selection accuracy
- Set stop conditions and budget guards so runs cannot loop indefinitely
- Reproduce reported agent failures and file each one with a minimal trace
- Present one experiment finding to the engineering team every month
Requirements
- Final year student or fresh graduate in engineering, science or a related field
- Python fundamentals: functions, dictionaries, virtual environments and reading a stack trace
- Curiosity about how LLMs decide, shown through a side project, notebook or blog post
- Willing to work from the Pune office two or three days a week
- Patience for debugging long logs without losing the thread
Skills
Benefits
- Competitive monthly stipend across the full six month term
- Weekly pairing with a senior applied AI engineer
- Ownership of a component that runs in production
- Completion certificate and a recommendation letter
- Conversion to a full time offer for interns who ship