As a Backend Engineer at the company, you will design and build a multi-agent Python pipeline that transforms unstructured exam papers (PDFs) into structured question–answer schemas used by our student-facing application. You will work with FastAPI, LLMs, OCR/Vision models, LangChain-style agent frameworks, and scalable cloud infrastructure.

This role sits at the intersection of GenAI engineering, backend systems design, and EdTech product development.

Key Responsibilities

1. Backend Engineering (FastAPI & Pipeline Architecture)

• Architect and implement a modular, multi-stage Python pipeline covering OCR, content extraction, parsing, and question structuring.

• Own, deploy, and maintain FastAPI microservices that integrate with LLMs, OCR engines, and internal data stores.

• Build reliable logging, monitoring, and cost-tracking systems across all pipeline components and agents.

2. GenAI, LLMs & Multi-Agent Systems

• Create multi-agent orchestration flows using frameworks like LangChain, LlamaIndex, or custom-built agent systems.

• Implement LLM-driven extraction, structuring, and reasoning workflows for exam content.

• Manage agent input/output schemas, reasoning traces, and message protocols.

• Ensure consistently structured outputs across a wide range of exam formats.

3. OCR & Computer Vision

• Integrate OCR and vision models for extracting text, diagrams, and images from PDFs.

• Handle multimodal input types such as equations, tables, graphs, and visual elements.

• Ensure reliable capture of mathematical notation and diagram-based questions.

4. Question Schema & Data Modelling

• Design scalable schemas for:

  • Questions and sub-questions

  • Answer keys

  • Diagrams, images, and mathematical expressions

  • Metadata supporting automated/LLM-based grading

• Ensure schema output supports both UI rendering and automated evaluation workflows.

5. Quality, Observability & Reliability

• Implement structured, timestamped logging at each pipeline stage.

• Build monitoring layers to track pipeline failures, false positives, and system health.

• Develop automated tests for multi-stage and multi-agent workflows.

Technical Requirements

Core Skills

• Strong proficiency in Python (3.9+)

• Experience building production systems using FastAPI

• Deep hands-on experience with LLMs, prompt engineering, and agentic workflows

• Familiarity with OCR engines (Tesseract, PaddleOCR, Vision-Language models such as GPT-5, Gemini 3 Pro, Qwen3-VL, etc.)

• Experience with PDF parsing libraries (pymupdf, pdfplumber, unstructured, etc.)

• Understanding of LangChain BaseMessage schemas or equivalent custom agent designs

• Ability to build and manage multi-stage pipelines with clean architecture

Software Engineering Best Practices

• Strong focus on modular design, documentation, and maintainability

• Experience implementing logging, monitoring, and analytics for backend systems

• Comfort working with cloud environments for scaling inference workloads

• Experience with CI/CD, unit testing, and integration testing

• Experience modelling structured and unstructured data

Bonus Skills

• Experience with:

  • Vector databases (Pinecone, Weaviate, Chroma)

  • Multi-agent orchestration frameworks

  • Educational content processing / assessment workflows

• Knowledge of math or science content (optional but helpful)

What You Will Build (Example Projects)

You will directly contribute to:

• Automated OCR → Question Schema → Answer Key pipelines

• LLM-based grading systems for handwritten or digital submissions

• Multi-agent reasoning systems with detailed trace logging

• Scalable ingestion frameworks capable of processing 10,000+ exam papers across regions and syllabi

• Multimodal processing pipelines for diagrams, charts, graphs, and mathematical expressions