Role & responsibilities

- Claude API & Anthropic SDK proficiency Hands-on experience with the Messages API, tool use / function calling, system prompt design, and model selection tradeoffs (Sonnet vs. Opus vs. Haiku). Familiarity with context window management and token budgeting.

- Agentic loop architecture — Ability to design reliable multi-step agent loops: tool orchestration, retry logic, error recovery, and knowing when to stop or escalate rather than loop indefinitely.

- Tool/MCP integration — Experience building and connecting tools (internal APIs, databases, external services) via Anthropic's tool use schema or MCP servers, including input validation and graceful failure handling.

- Prompt engineering & evaluation — Skilled at structured prompting (system prompts, few-shot examples, XML tagging), and building prompt eval harnesses to measure output quality, regression-test changes, and tune instructions systematically.

- Observability & auditability — Knows how to log full agent traces (inputs, tool calls, intermediate outputs, final responses) in a structured, queryable format.

Experience with tools like LangSmith, Braintrust, Helicone, or custom tracing pipelines.

- Measurement & KPI design — Can define and instrument meaningful agent metrics: task completion rate, tool call accuracy, hallucination rate, latency per step, cost per run, and human-in-the-loop escalation rate. Connects agent telemetry to business outcomes.

- Human-in-the-loop & guardrails — Understands when to inject human review checkpoints, how to design approval gates for high-stakes actions, and how to implement input/output guardrails (content filtering, schema validation, confidence thresholds).

- Cost & latency optimization — Experience profiling and reducing inference costs through prompt caching, batching, streaming, and appropriate model tiering — without sacrificing reliability.





- Security & data handling — Awareness of prompt injection risks, credential/secret hygiene in agentic contexts, PII handling, and least-privilege design when agents have access to real systems or external APIs.

- Software engineering fundamentals — Robust async Python (or TypeScript), testing discipline (unit + integration tests for agent components), CI/CD, and the ability to decompose complex agent systems into maintainable, modular code.

AI / LLM

- Hands-on experience with LLM APIs (OpenAI, Anthropic, Cohere, or similar)

- Experience building agentic systems (tool use, memory, multi-step reasoning)

- Familiarity with prompt engineering techniques (chain-of-thought, few-shot, RAG)

- Understanding of fine-tuning and model evaluation concepts

Backend (Python)

- Strong proficiency in Python 3.x

- Experience with FastAPI or Django / Flask

- Working knowledge of SQL and NoSQL databases (PostgreSQL, MongoDB, Redis)

- Familiarity with async programming and background task queues (Celery, RQ)

- Experience with Docker and deploying to cloud platforms (AWS, GCP, or Azure)

Frontend (React)

- Strong proficiency in React.js and modern JavaScript (ES6+)

- Experience with TypeScript

- Familiarity with state management (Redux, Zustand, or Context API)

- Ability to build streaming UI for LLM outputs (token-by-token rendering)

- Basic understanding of UX principles for AI interfaces

General

- Experience with Git and collaborative development workflows

- Comfort working in fast-paced, ambiguous environments

- Strong problem-solving and communication skills

Nice to Have

- Certified Claude Architect – Foundations (CCA-F)

- Experience with Claude Code, Cursor, or other AI-assisted development tools

- Contributions to open-source AI projects

- Experience with multi-agent frameworks (AutoGen, CrewAI, LangGraph)

- Knowledge of MLOps practices and model deployment (MLflow, Weights & Biases)

- Familiarity with WebSockets for real-time AI streaming

- Experience with Kubernetes or serverless architectures

📌 Generative AI Fullstack developer (Bengaluru)
🏢 Bounteous
📍 Bengaluru