AI/ML Engineer — LLMs, Memory & Personalization
Role Overview
We are looking for an AI/ML Engineer to build and evolve MyNaksh's AI intelligence layer, focusing on conversational AI, long-term memory, personalization, and domain-specific model development for astrological reasoning.
Key Responsibilities
- Build conversational AI systems with contextual understanding, long-term memory, and personalized responses.
- Design and optimize RAG pipelines using embeddings, vector search, structured data, and knowledge graphs.
- Develop systems for user profiling, memory extraction, retrieval, ranking, and context selection.
- Integrate, deploy, and fine-tune open-source LLMs and develop specialized SLMs for astrological reasoning.
- Build domain-specific training datasets and pipelines for fine-tuning, model adaptation, and continuous improvement.
- Develop LLM orchestration and lightweight ML systems for classification, ranking, and personalization.
- Build evaluation frameworks to measure retrieval quality, memory accuracy, astrological reasoning, and response consistency.
- Optimize model performance, inference costs, latency, and reliability in production.
- Own end-to-end AI engineering, from experimentation and prototyping to deployment, monitoring, and iteration.
Required Skills & Experience
- Strong Python programming and experience building production-grade ML or AI systems.
- Hands-on experience with LLMs, open-source models, embeddings, vector databases, and RAG.
- Understanding of LLM fine-tuning, model evaluation, dataset preparation, and inference optimization.
- Experience building retrieval, memory, or personalization systems.
- Familiarity with LLM orchestration, prompt engineering, and multi-step AI pipelines.
- Experience with model serving, APIs, testing, monitoring, and production deployment.
- Strong problem-solving skills and the ability to evaluate AI improvements using measurable outcomes.
Good to Have
- Experience fine-tuning or deploying small language models (SLMs) using techniques such as LoRA/QLoRA.
- Experience with knowledge graphs, graph-based retrieval, or Neo4j.
- Experience with synthetic data generation, domain-specific training datasets, and automated evaluation pipelines.
- Experience in recommendation systems, ranking models, or lightweight ML classifiers.
- Familiarity with GPU inference optimization and self-hosted open-source models.
- Experience in astrology, symbolic reasoning, or other specialized domain-reasoning systems.
What You'll Build at MyNaksh
- Personalized AI experiences powered by conversational context and long-term user memory.
- Retrieval and intelligence systems that make responses more relevant and consistent over time.
- Specialized models and evaluation pipelines that improve astrological reasoning using domain-specific data.
- Scalable AI infrastructure that balances response quality, latency, and cost.
