Saigon Technology Solutions

MUST HAVE SKILLS:

• Have programming skills in one of these languages: Python...
• At least 3 years of experience working with AI/ML projects as a Data Engineer, Research Engineer, or Software Engineer.
• Have experience with one of ML/DL Frameworks: Tensorflow, Pytorch, scikit-learn, Pandas, Langchain, Langraph, LlamaIndex…
• Have experience working with OpenAI, Claude, Gemini, any LLM... to build Agentic Solution, Workflows, or advanced RAG systems, using MCP, Tool call, handling Agent Memory/Context.
• Have experience with one of chatbot building frameworks or services, Agentic System.
• Familiar with OCR algorithms or services: OpenCV, Tesseract, Textract (AWS), Google Cloud Vision, PaddleOCR.
• Experience with common development tools: Linux, GPU server, Google Colab, Jupyter, Git, Docker, vLLM, Ollama.
• Understanding of LLM security risks and mitigation techniques: prompt injection, jailbreaking, data/PII leakage, and adversarial inputs.
• Experience implementing guardrails for LLM/agentic systems (e.g., input/output validation, content moderation, PII redaction, rate limiting, tool-call permission scoping) using frameworks/services such as Guardrails AI, NeMo Guardrails, Llama Guard, Presidio, or custom rule-based/classifier approaches.
• Good English proficiency and communication skills.

NICE TO HAVE

• Experience with backend web development (Flask, Django, FastAPI)
• Experience with SQL database (PostgresSQL, MySQL), NoSQL database (MongoDB), Vector database (Qdrant, Chroma, OpenSearch, PgVector...)
• Experience with building APIs and services using REST, SOAP, SSE, Websocket etc.
• Experience in MLOps system: MLflow, DVC, Feature Store, Airflow, CI/CD...
• Experience in Generative AI, Large Language Model, Prompt/Context Engineer, Diffusion model, HuggingFace, Ollama, LlamaCPP
• Familiarity with responsible AI practices: bias/fairness evaluation, red-teaming LLM applications, and audit logging for compliance.
• Knowledge of secure API design and authentication/authorization patterns (OAuth2, API key scoping, least-privilege access) for AI service endpoints.