Kishan

Senior machine learning engineer

AIHR • Fulltime

May 2023 - Present • 3 yrs

● LLM Systems & MLOps: Deployed on-premise LLM and VLM on Kubernetes using vLLM inference engine to scale～500 concurrent users with low-latency performance. ● Implemented GPU and inference monitoring (Prometheus, Grafana) to improve system observability and reduce production downtime. ● Generative AI / RAG: Designed RAG based chatbot to recommend procurement procedure under EU and German legal regulations. ● Utilized domain-specific embedding model and retrieval tuning to improve retrieval accuracy and reduce manual decision-making time by ～40-50%. ● Time Series & Analytics: Developed predictive anomaly detection pipeline for solar plant performance monitoring and power loss identification. ● Reduced incident detection time by ～30% to allow faster mitigation and minimize revenue loss. ● Computer Vision: Implemented SAM 2 assisted annotation workflow and trained YOLOv11 model to classify and quantify alive vs dead nematodes in biological agricultural research. ● Reduced manual nematode counting time by ～80% and improved counting accuracy to ～90%.

Senior Software Engineer

Cloudtheapp • Fulltime

Dec 2020 - Apr 2023 • 2 yrs 4 mos

• Built Python-based backend systems on GCP powering automated merchant lending decisions — approvals, denials, re-approvals, and loan updates at scale using Airflow-orchestrated pipelines and BigQuery • Created and optimized lending operations features using dbt, improving downstream model performance, reporting consistency, and analytics readiness across credit workflows. • Built AI-powered merchant scoring pipelines that improved lending decision accuracy by 25% and increased lending processing efficiency by 40%, supporting faster credit approvals at scale. • Designed and maintained automated decisioning systems responsible for approving, denying, re-approving, and updating 10K+ merchant loans weekly with high reliability and low manual intervention. • Developed an automated experiment deployment platform for data scientists, reducing model rollout time from weeks to days, enabling 3x faster experimentation and more precise capital allocation across lending products.