AI Engineer building production agentic systems. MS in AI from Northeastern, graduating May 2025.
At Vizit, I built Project Spark: an agentic system that orchestrates multiple analysis modules (benchmarking, clustering, vision-language reasoning) to diagnose and fix low-performing e-commerce images. Lifted visual appeal scores by 30% across 20+ global brands. Also built a 5M+ scale semantic search pipeline with CLIP embeddings, drift monitoring, and quantized inference.
Previously at AG Mednet, I built Judi-Insights: agents that wrote complex SQL, queried production DBs, and delivered clinical trial reports with <5% hallucination. Modular RAG with custom rerankers on vector DB.
Side projects: Kube-Agent (LLM-based SRE for Kubernetes via natural language), PalursRAG (attention-based dynamic chunk reordering).
Research: multimodal attention on 150K video dataset, Digital Twin personalization with LoRA adapters. Published at IEEE AIxSET 2025, Applied Sciences. Patent holder.
Looking for: AI/ML Engineer, Applied Scientist, or Agent Engineer roles. Interested in agentic systems, multimodal AI, and production ML. I ship fast and own problems end-to-end.
Technologies: Python, PyTorch, LangChain/LangGraph, RAG, Agentic AI, CLIP, LoRA/QLoRA, Quantization, AWS (SageMaker, Bedrock, EKS), Docker, Kubernetes, Vector DBs (FAISS, Pinecone, Chroma), PostgreSQL, FastAPI
Email: pkompally@gmail.com GitHub: https://github.com/SteveMama
AI Engineer building production agentic systems. MS in AI from Northeastern, graduating May 2025.
At Vizit, I built Project Spark: an agentic system that orchestrates multiple analysis modules (benchmarking, clustering, vision-language reasoning) to diagnose and fix low-performing e-commerce images. Lifted visual appeal scores by 30% across 20+ global brands. Also built a 5M+ scale semantic search pipeline with CLIP embeddings, drift monitoring, and quantized inference.
Previously at AG Mednet, I built Judi-Insights: agents that wrote complex SQL, queried production DBs, and delivered clinical trial reports with <5% hallucination. Modular RAG with custom rerankers on vector DB.
Side projects: Kube-Agent (LLM-based SRE for Kubernetes via natural language), PalursRAG (attention-based dynamic chunk reordering).
Research: multimodal attention on 150K video dataset, Digital Twin personalization with LoRA adapters. Published at IEEE AIxSET 2025, Applied Sciences. Patent holder.
Looking for: AI/ML Engineer, Applied Scientist, or Agent Engineer roles. Interested in agentic systems, multimodal AI, and production ML. I ship fast and own problems end-to-end.