Ram Dheeraj Kamarajugadda
About
Detail
North Carolina, United States
Contact Ram regarding:
work
Full-time jobs
Flexible work
id_card
Internships
groups
Networking
Résumé
Jobs
verified_user
0% verified
- UAI EngineerUSA National Phenology Network (NPN)Feb 2024 - Current (2 years 3 months)Designed and deployed multi-model GenAI assistants using GPT-4, Gemini, Mistral, Groq, automating scientific querying for ~2,000 monthly researcher interactions.Built production-grade RAG pipelines (LangChain + FAISS/ChromaDB + HuggingFace), reducing ecological data retrieval from days to <40 seconds.Engineered Python + SQL event pipelines ingesting 7,000+ monthly events, producing ML-ready datasets powering downstream Pandas/NumPy transformations.Developed automated LLM monitoring systems for latency, drift, hallucinations, and endpoint failures, improving anomaly detection speed by 2.5x.Implemented embedding pipelines using HuggingFace Transformers + AWS Lambda, generating 100K+ embeddings/week for semantic search and vector retrieval.Bui
Software Engineer, ML/AIJP Morgan Chase & Co.Mar 2021 - Jul 2023 (2 years 5 months)Engineered a production NLP incident-triage engine using Python and embeddings (spaCy + Sentence Transformers) that recommended resolutions from historical tickets, improving response efficiency and reducing handling time by 40%Built automated anomaly-detection intelligence using Splunk, Python heuristics, and ML-driven pattern analysis across 25+ middleware applications, preventing outages and generating $60K/year in early-issue avoidance.Designed high-reliability monitoring dashboards using Python, MySQL, Tableau, and AWS-hosted data sources to track middleware health, reducing manual investigation effort by 50%.Automated deployment QA by building a Python ETL pipeline with Jenkins CI/CD + AWS S3 as the storage layer for validation logs,- NMachine Learning EngineerNavana TechJan 2018 - Feb 2021 (3 years 2 months)Developed ML models for classification, ranking, and semantic similarity using TensorFlow, Scikit-learn, and Transformers, supporting 25K+ daily inferences with sub-250 ms latency.Built vector search pipelines with FAISS + custom embedding stores, reducing lookup latency from 420 ms → 240 ms and enabling 18K+ additional successful retrievals/month.Created end-to-end training pipelines for structured and unstructured data using Python, Pandas, NumPy, and tokenization workflows, reducing failed training runs by 150–180 per quarter.Performed large-scale conversational analytics using NLP preprocessing, embedding similarity checks, and drift analysis, reducing hallucination-type errors by ~1,300 per month.Authored ML documentation, schema maps,
Education
verified_user
0% verified
- UMaster of Science (M.S.) in Data ScienceUniversity of Arizona, Tucson, AZAug 2023 - May 2025 (1 year 10 months)