Required skills

AreaRequirement

LanguagesPython at expert level. TypeScript or JavaScript at working level.
ML stackPyTorch, Hugging Face Transformers, TRL, PEFT, Accelerate, Datasets
TrainingSFT, LoRA and QLoRA, distributed training, dataset curation
AlignmentHands-on work in at least one of RLHF, DPO, GRPO, or RLVR, with measured results
AgentsTool calling, MCP, agent loops, context management, agent evaluation
InferencevLLM, TGI, or llama.cpp. Quantisation and latency tuning.
BackendFastAPI, async Python, Pydantic, PostgreSQL, Redis
DataEmbeddings, vector databases, RAG pipelines, chunking strategy
OperationsDocker, Git, Linux, CI/CD, GPU basics (CUDA, VRAM planning)
EvaluationBenchmarks, custom eval sets, LLM as judge, regression tests

Good to have

  • Federated learning, differential privacy, or other privacy preserving training methods.
  • Custom CUDA or Triton kernels.
  • Model distillation and model merging.
  • Guardrails, red teaming, and jailbreak testing.
  • Multimodal models: vision, speech, or document understanding.
  • Open source contributions to any part of this stack.
  • Experience with on-premise or air-gapped deployment.