Prayas AI is an EdTech platform revolutionizing exam preparation for UPSC, BPSC, and UPPCS aspirants. Our core product provides "Instant Feedback" on handwritten answer sheets. We are looking for a pragmatic engineer to join our team and help build the intelligence pipelinefrom orchestrating workflows to generating high-quality, syllabus-aligned critiques.

The Role :

You will be a core member of the engineering team working on the "Evaluation Engine." You won't just write prompts; you will collaborate to build the infrastructure that ensures our AI is accurate, reliable, and cost-effective. You will combine GenAI (to grade answers), RAG (to know the facts), and Workflow Automation (n8n) to deliver a seamless experience.

Key Responsibilities :

1. GenAI, Prompting & Tuning :

- Prompt Architecture : Design sophisticated prompt chains (Chain-of-Thought, Few-Shot) for the Gemini API to evaluate subjective answers.

- Model Fine-Tuning : Identify when prompting hits a limit and lead initiatives to Fine-Tune models based on requirements.

- Structured Output : Ensure the LLM outputs strict, reliable JSON schemas (using Pydantic or Geminis JSON mode) so that the data integrates seamlessly with our Frontend UI.

- Knowledge Base (RAG) : Work on the retrieval system (using pgvector or similar) that fetches the correct "Model Answers" and "Current Affairs" to ground the AI's feedback in fact.

2. Computer Vision & OCR :

- OCR Optimization : Fine-tune the pipeline to handle mixed-language scripts (English + Hindi terms) common in UPSC answers, ensuring accurate text extraction from student uploads.

3. Engineering & Automation :

- Workflow Orchestration (n8n) : Build and maintain robust n8n workflows to manage the API lifecycle. You should know when to use n8n for speed and when to switch to Python for complexity.

- Backend API : Develop FastAPI endpoints for custom logic that n8n cannot handle.

- System Reliability & Cost : Monitor API token usage and implement caching strategies to keep

operational costs low without sacrificing quality.

- Version Control : Maintain clean Git history for Python services and implement backup/versioning strategies for n8n workflows.

4. Product & Collaboration :

- Translating Subject Expertise : Work closely with our Academic Team (UPSC Mentors). Your job is to translate their subjective grading intuition ("This answer lacks flow") into concrete engineering logic.

- Automated Evaluations ("Evals") : Build automated test pipelines to quantitatively measure the accuracy of the AI against human-graded papers.

Required Skills :

- Core : Strong proficiency in Python and SQL (PostgreSQL).

- GenAI Native : Deep experience with LLMs (Gemini/OpenAI), Prompt Engineering, and RAG architectures.

- Model Tuning : Understanding of when and how to Fine-Tune models vs. using RAG. Experience preparing training datasets for Supervised Fine-Tuning (SFT) is a big plus.

- Data Structure : Experience enforcing JSON schemas on LLM outputs (e.g., using Instructor, Pydantic, or LangChain).

- Automation : Proven experience with n8n (or similar low-code tools) for production workflows.

- Computer Vision : Familiarity with OCR APIs (Textract, Google Vision) and handling text extraction challenges.

Nice to Have :

- Experience building "Eval" frameworks (using tools like LangSmith, DeepEval, or custom scripts).

- Background in EdTech or experience with the UPSC ecosystem.

The job is for:

Women candidates preferred