Full Stack Developer (AI/ML & Web Applications)

Full Stack Developer (AI/ML & Web Applications)

Role Summary

The Full Stack Developer will design, build and deploy secure, scalable web applications with integrated AI and machine learning capabilities. The role combines strong full-stack engineering across frontend, backend, data and cloud layers with hands-on competence as an AI/ML engineer: developing, fine-tuning, optimising and serving machine learning, deep learning and large language models, and integrating them across commercial, cloud and self-hosted model providers.

Educational Qualification

• B.Tech./B.E., M.Tech., MCA or M.Sc. in Computer Science, Information Technology, AI/ML, Data Science or a related discipline (Must have)
• Professional certification in machine learning, deep learning or cloud platforms is desirable but not mandatory; demonstrated work, as listed under Minimum Evidence at Screening, shall carry greater weight
• Candidates with strong verifiable work (production applications, published models or significant open-source contributions) will be considered in lieu of formal qualification

Experience

• 4+ years of professional software development, including full-stack web applications (5+ years preferred); minimum 2 years developing, fine-tuning and deploying AI/ML models and LLM applications in production
• Demonstrable end-to-end ownership of at least one production web application, covering frontend, backend, database, deployment and monitoring
• Demonstrable experience training or fine-tuning at least one model for NLP, computer vision or speech and deploying it for end users, with documented evaluation
• Demonstrable experience building at least one LLM application (such as a retrieval-augmented assistant, a document-processing workflow or an agent) integrated with more than one model provider
• Hands-on experience with Python and TypeScript in production, and with React-based frontends
• Experience deploying containerised applications on cloud platforms through CI/CD; experience with GPU-based inference is desirable
• Experience in government, financial services, healthcare or other regulated sectors, or with Indian-language applications, is a plus

Key Responsibilities

Application Development

• Design and develop responsive, accessible web applications using React and Next.js with TypeScript
• Develop backend services and APIs in Python (FastAPI) and Node.js, with clear contracts (OpenAPI or GraphQL) and versioning
• Design data models and optimise queries across relational and NoSQL databases, and implement caching and background job processing
• Translate functional requirements into well-structured, maintainable and production-ready code

AI/ML Model Development

• Develop, train and fine-tune machine learning and deep learning models for use cases such as text classification and summarisation, entity extraction, OCR and document understanding, image matching and object detection, and speech recognition
• Fine-tune open-weight LLMs and embedding models on domain data (LoRA or QLoRA), and distil larger models into smaller, cost-efficient models for production
• Implement data preprocessing, feature engineering and data augmentation pipelines; run tracked experiments to improve accuracy and latency; and perform error analysis to improve model stability
• Optimise models for inference through quantisation, pruning, distillation and export to ONNX or TensorRT, and benchmark latency and throughput on CPU and GPU
• Serve models in production with batching, autoscaling and GPU utilisation monitoring

LLM Integration Across Providers

• Integrate LLMs in a provider-agnostic manner across commercial APIs, cloud AI platforms, Indian-language providers and self-hosted open-weight models, through an abstraction or gateway layer, so that models can be switched, compared or combined without application changes
• Implement model routing, caching, rate limiting and fallbacks across providers, and track quality, cost and latency for each provider
• Build LLM-powered features, including streaming conversational interfaces, retrieval-augmented generation with source citations, structured extraction, summarisation and tool-calling agents using the Model Context Protocol (MCP)
• Implement retrieval pipelines covering document ingestion and parsing, chunking, embeddings, hybrid vector search and re-ranking

Evaluation & Model Lifecycle

• Evaluate models with task-appropriate metrics (such as F1, AUC, BLEU, ROUGE, WER, CER and mAP), and evaluate LLM features with golden datasets and LLM-based evaluation calibrated against human labels
• Compare models and providers on accuracy, cost and latency before adoption, and run evaluation suites in CI as release gates
• Version, register and monitor models with the data science and MLOps teams, and build feedback loops that capture user corrections for retraining

Security, Privacy & Responsible AI

• Implement authentication and authorisation (OAuth 2.0, OpenID Connect and role-based access control), and secure handling of APIs, secrets and personal data
• Apply OWASP Top 10 and OWASP Top 10 for LLM Applications controls, including defences against prompt injection and data leakage
• Ensure model outputs comply with Responsible AI guardrails and applicable data protection law, including for data sent to external model providers, and design human review for consequential AI outputs

Quality, Deployment & Operations

• Practise test-driven development with unit, integration and end-to-end tests, and maintain adequate coverage
• Containerise applications and models, and manage deployments through CI/CD pipelines on Kubernetes or serverless platforms, using infrastructure as code
• Instrument applications with logging, metrics and tracing, including LLM traces, token usage, cost and GPU utilisation, and participate in incident resolution
• Ensure accessibility compliance (WCAG 2.2 and GIGW) and support for Indian languages in user interfaces

Collaboration & Mentoring

• Participate in Agile delivery, code reviews and architecture discussions, and support integration of AI components into existing applications
• Mentor junior developers, and document designs, models, APIs and operational runbooks

Technical Competencies

• Frontend: TypeScript, React, Next.js (App Router and server components), Tailwind CSS, component libraries (shadcn/ui or equivalent), TanStack Query; accessibility and internationalisation
• Backend: Python (FastAPI, Pydantic, asyncio) and Node.js (NestJS, Fastify or Express); REST with OpenAPI, GraphQL; WebSockets and server-sent events for streaming; background jobs (Celery or BullMQ)
• Databases & Caching: PostgreSQL (with pgvector), MongoDB, Redis; ORMs (SQLAlchemy, Prisma or Drizzle); schema migrations and query optimisation
• AI/ML Frameworks: Python (expert level), PyTorch, TensorFlow or Keras, Hugging Face Transformers, PEFT and TRL, scikit-learn and XGBoost; OpenCV and spaCy; pandas and NumPy
• Model Optimisation & Serving: quantisation (GPTQ, AWQ or bitsandbytes), pruning and distillation; ONNX Runtime and TensorRT; vLLM, SGLang, NVIDIA Triton Inference Server or TorchServe; GPU inference and working knowledge of CUDA; distributed training (PyTorch FSDP or DeepSpeed) desirable
• Vision, Document, Speech & Indic AI: object detection and image-matching models; vision-language OCR and document parsing (Docling or equivalent); speech recognition (IndicConformer, Whisper); Indic NLP and translation (AI4Bharat models, IndicTrans2, Bhashini)
• LLM Providers & Platforms: OpenAI, Anthropic, Google Gemini, Azure AI Foundry, AWS Bedrock, Google Vertex AI and Sarvam; self-hosted open-weight models (Llama, Qwen, Mistral or Gemma) via vLLM or Ollama; provider gateways (LiteLLM or equivalent) and OpenAI-compatible APIs
• AI Application Frameworks: LangGraph, LlamaIndex or equivalent multi-provider frameworks; structured outputs, tool calling and the Model Context Protocol (MCP)
• Retrieval & Vector Search: embeddings, chunking strategies, hybrid search and re-ranking; vector databases (pgvector, Qdrant, Milvus or Weaviate)
• Evaluation & MLOps: F1, AUC, BLEU, ROUGE, WER, CER and mAP; promptfoo, DeepEval or RAGAS; MLflow or Weights & Biases, DVC and model registries; Langfuse or Arize Phoenix for LLM tracing
• Security: OAuth 2.0, OpenID Connect, JWT, Keycloak or equivalent identity providers; OWASP Top 10 and OWASP Top 10 for LLM Applications; secrets management; guardrails for LLM inputs and outputs
• Testing, DevOps & Cloud: pytest, Vitest or Jest, Playwright; Git, GitHub Actions or GitLab CI, Docker, Kubernetes and Helm, Terraform; AWS, Azure or GCP including GPU instances; OpenTelemetry, Prometheus and Grafana
• Data Pipelines (desirable): Airflow or Dagster; Kafka or other message queues