George Town, Malaysia · hace 5 meses
Desarrollar y ampliar servicios backend y flujos LLM con estado para una plataforma de IA agéntica, incluidos los sistemas de recuperación y puntuación. Depurar problemas de producción usando logs y trazas.
Overview We are hiring a Mid-to-Senior Full Stack Developer to build and extend an Agentic AI platform with strong focus on LLM workflows, orchestration, and backend systems. This role requires a highly independent engineer who can quickly understand existing codebases and deliver production-quality enhancements with minimal guidance. Core Hiring Bar (Non-Negotiable) • Ability to independently read, understand, and modify moderately complex Python modules (~300+ lines) • Comfortable filling knowledge gaps through documentation and reasoning (not dependent on AI-assisted coding tools) • Demonstrated ability to interpret existing systems and implement changes with minimal onboarding Key Responsibilities • Develop and enhance backend services and agentic workflows for the AI platform • Build stateful, multi-step LLM pipelines and orchestration logic • Design, optimize, and maintain retrieval and scoring systems • Debug production issues using logs, traces, and system behavior • Collaborate across teams to deliver scalable and reliable AI-driven solutions • Implement incremental changes with strong testing and validation practices Technical Requirements Python (Senior Level) • Writes clean, idiomatic Python using: • Type hints, dataclasses, Pydantic • Generators and context managers • Strong understanding of: • Async/await and concurrency models • Proficient in Python standard libraries (e.g., pathlib, json, re, collections) • Able to modify existing complex systems independently Backend Development (FastAPI) • Experience building and extending FastAPI services • Strong understanding of: • Request lifecycle • Dependency injection and middleware • Multi-worker deployments and shared state (e.g., Redis) • Able to diagnose issues using logs and traces (minimal debugger reliance) LLM Engineering (Applied) • Experience building production-grade LLM workflows • Strong in: • Deterministic prompt design (structured outputs, low/no temperature) • Handling failure modes (timeouts, malformed outputs) • Understanding of RAG systems: • Chunking, embeddings, similarity scoring Workflow Orchestration (LangGraph or Equivalent) • Experience with stateful orchestration frameworks preferred • Must be able to quickly: • Learn graph/state concepts • Implement multi-step workflows within 1–2 weeks Retrieval & Scoring Systems • Experience with ranking/scoring methods (e.g., BM25, hybrid search) • Ability to tune: • Thresholds, weighting, precision vs recall trade-offs • Capable of building realistic test datasets Diagnostics & Log Processing • Familiar with log ingestion and analysis pipelines • Understands: • Chunking strategies • Pattern extraction vs LLM reasoning • When to use deterministic vs AI-based parsing Infrastructure & Runtime • Hands-on experience with: • Docker / Docker Compose (volumes, dependencies, health checks) • Debugging container runtime issues • Working knowledge of: • Redis (basic operations, TTL, persistence) • Enterprise networking concepts (e.g., proxies) Frontend (Working-Level) • Ability to work with HTML + Vanilla JavaScript • Comfortable with: • DOM manipulation • Fetch APIs and event handling • Able to implement UI changes from requirements (no design dependency) Work Style Expectations • Strong code-first discipline (understands design before coding) • Built-in focus on: • Testing and validation • Log-driven debugging • Writes clean, incremental changes with clear commits • Performs self-review against acceptance criteria • Asks focused, implementation-driven questions Experience & Seniority • Level: Mid to Senior Engineer • Experience: • ~4–8 years in Python development • Proven track record delivering production systems • Experience with LLM-based or AI platforms preferred Ideal Candidate Profile • Highly independent problem-solver • Strong systems thinker (not just feature coder) • Comfortable working in low-AI-assist environment • Bias toward execution, debugging, and delivery quality
Publicada el 5 jun 2026
