Full-Stack Applied AI Engineer (Social Commerce & LLM Agents)

Full-Stack Applied AI Engineer (Social Commerce & LLM Agents)Location: AUPP Tech Center, Phnom Penh, Cambodia

Employment Type: Full-Time, OnsiteCompensation: Competitive startup salary based on technical portfolio and experienceApplication Contact: info@tobsar.comAbout TobSarTobSar is an AI sales agent platform purpose-built for social commerce businesses.

We empower merchants to manage customer conversations, automate sales and order workflows, and improve daily operations across messaging and social channels.In Southeast Asia’s chat-driven economy, transactions take place inside conversation windows on Facebook Messenger, Telegram, and TikTok. TobSar develops intelligent, reliable AI Sales Agents that respond instantly, track orders, and convert chats into revenue. Our core technical focus is building frugal, sub-second systems tailored to how Cambodians communicate online—supporting colloquial Khmer, code-switching, local slang, and real-world sales patterns.Role OverviewWe are hiring a product-driven Full-Stack Applied AI Engineer to join our founding technical team onsite at the AUPP Tech Center in Phnom Penh. In this role, you will take features from architectural design to deployment across our conversational agent engine, webhook integrations, and merchant portal.You will work directly with foundational LLM models, tool-calling pipelines, structured generation, and real-time messaging APIs, creating an end-to-end platform that local sellers can depend on 24/7.Key ResponsibilitiesConversational Agent Architecture: Design, test, and optimize stateful multi-turn conversational agents, prompt chains, and tool-use pipelines for automated order intake, pricing inquiries, and product recommendation.Khmer NLP & Conversational Edge Cases: Tackle challenges around mixed-language chat, colloquial Khmer slang, typos, and informal speech to ensure high intent detection accuracy.Full-Stack Development: Build and maintain scalable backend APIs (FastAPI/Django/Node) and responsive merchant dashboards (React/Next.js, Tailwind CSS) for live supervision, human takeover, and order tracking.Messaging & Webhook Pipelines: Integrate and maintain high-throughput webhooks with social messaging platforms (Facebook Graph API/Messenger, Telegram Bot API, TikTok) and order databases.Latency, Cost & Reliability Optimization: Implement efficient caching layers, prompt caching, token budgets, and fallback routines to achieve sub-second response times economically.Guardrails & Evaluation: Establish automated evaluation benchmarks, hallucination guards, and deterministic verification for sensitive transactions (stock validation, payment confirmation).Requirements & Technical QualificationsFull-Stack Engineering: Hands-on experience developing web applications end-to-end using modern frameworks (e.g., Python FastAPI/Django, TypeScript/JavaScript, React/Next.js, Tailwind CSS).Applied AI & LLM Orchestration: Practical experience working with LLM APIs, function calling / structured JSON outputs, prompt engineering, and RAG/vector retrieval.Database & Systems Design: Working knowledge of relational databases (PostgreSQL/MySQL), Redis caching/queues, and asynchronous event-driven architectures.Deployment & DevOps: Comfort deploying and maintaining services using Docker, Nginx, and cloud hosting infrastructure.Language & Communication: Professional proficiency in written and spoken English and Khmer.Mindset: Resourceful, proactive, and committed to shipping reliable products in an early-stage startup environment.Role Details & PathDescription Work ArrangementFull-time, OnsiteLocationAUPP Tech Center, Phnom Penh, CambodiaTeam StageEarly Core Technical TeamGrowth PathDirect pathway toward engineering leadership (Tech Lead / Head of Engineering) as the team expands.How to ApplyInterested candidates should submit the following to info@tobsar.com with the subject line "Application: Full-Stack Applied AI Engineer — [Your Name]":Resume / CV or detailed LinkedIn profile.GitHub profile or portfolio links showing past web apps, APIs, or AI/LLM integration projects.Short Technical Question: "How would you architect an automated fallback mechanism when an AI sales agent receives an ambiguous customer input (e.g., a blurry payment screenshot or heavy Khmer slang) to prevent incorrect transactions?"