Website:
nasugroup.com
Job details:
Join a team building GenAI and agent-based applications for financial institutions — from understanding
complex systems to accelerating engineering delivery. You’ll develop LLM-powered agents and retrieval
systems that reason over large codebases and enterprise data to produce real engineering outcomes.
Roles & Responsibilities
• Design, build and ship LLM-powered and agentic applications — multi-step, tool-using, reliable.
• Build retrieval-augmented (RAG) systems over large, messy enterprise data — source code,
documentation, schemas.
• Work with vector databases and knowledge graphs for retrieval and reasoning.
• Engineer prompts, evaluations and guardrails; measure and systematically improve output quality.
• Integrate and route across multiple LLM providers; optimise for cost, latency, and on-prem / air-gapped
constraints.
• Build backend services and automation that turn model output into auditable, production-grade
artefacts.
Mindset & problem-solving
• First-principles thinker — decomposes ambiguous, open-ended problems and reasons from
fundamentals rather than reaching for the nearest template.
• Inventive solutioner — proposes novel approaches, challenges assumptions, and weighs trade-offs to
find the best answer, not just a working one.
• Bias to prototype — experiments quickly, learns from what the models and data actually do, and
iterates.
• Strong analytical ability and genuine curiosity; comfortable when the path isn’t defined.
Must-have
• Strong Python.
• Hands-on experience building LLM applications — agents, tool use, RAG.
• Vector search (pgvector / FAISS / similar); knowledge graphs (Neo4j / Cypher) a plus.
• Prompt engineering plus systematic LLM evaluation.
• FastAPI / async services, containers (Docker / Podman), Git.
Nice-to-have
• Familiarity with any Agent Development Kit / framework (e.g. Google ADK, OpenAI Agents SDK,
LangGraph, CrewAI, AutoGen).
• Serena or similar semantic code-understanding / coding-agent toolkits (LSP- / MCP-based) for
reasoning over and navigating large codebases.
• Model gateways (LiteLLM), MCP, orchestration libraries (LangChain / LlamaIndex).
• Code analysis / parsing / AST or static-analysis work.
• Delivery in regulated / on-prem / air-gapped environments; financial-services exposure.
Click on Apply to know more.
