Posted 08 July, 2026
Research Engineer
Diverse Lynx India
Bengaluru, Karnataka, IN
Full Time
Reference: 26-01138-575-2
Research Engineer | Title: AI Research Engineer
The Research Engineer will be involved in AI feature development and architect scalable ML systems. The position focuses on building LLM, CV, and multimodal ML pipelines, deploying models in production, and ensuring reliability and cost efficiency.
Key Responsibilities
- Lead design and development of core AI features - from data ingestion to real-time inference for production products.
- Architect ML systems and services that are scalable, cost-efficient, and observable.
- Build LLM, CV, or multimodal (ML) training pipelines (fine-tuning, adapters, retrieval, and evaluation) depending on product needs.
- Define model evaluation frameworks (offline metrics + live A/B + user feedback loops).
- Collaborate with Software, Data, and Product teams to design features powered by ML.
- Deploy and monitor models using containerized microservices (ECS/K8s), ensure low-latency inference and reproducibility.
- Own incident response and postmortems for AI systems, improve reliability and reduce MTTR.
- Optimize training/inference cost (batching, quantization, mixed precision, GPU scheduling).
- Education: Bachelors/Master's from a top-tier institute (IIT/Tier-1 etc.) in Computer Science, AI, or related field.
- 1–5+ years in applied ML/AI roles
- Expert in Python, strong in at least one deep-learning framework (PyTorch/TensorFlow).
- Experience with end-to-end ML pipelines (data prep, training, evaluation, deployment, monitoring).
- Proven success shipping ML-powered products not just models to real users.
- Hands-on with MLOps tooling (MLflow, Weights & Biases, DVC, Airflow, Prefect, etc.).
- Knowledge of containerized deployments (Docker, ECS, K8s) and CI/CD for ML.
- Strong fundamentals in statistics, experimentation, and interpreting real-world feedback.
- Experience optimizing/operating GPU inference (ONNX, TensorRT, mixed precision, batching).
- Familiar with vector databases, RAG pipelines, and LLM fine-tuning/adapters (LoRA/QLoRA).
- Exposure to observability stacks (Prometheus/Grafana/OpenTelemetry) and production logging/metrics.
