“Working at the intersection of GenAI and information retrieval, I specialize in LLM fine-tuning, RAG, relevance ranking, search summarization, and LLM-driven evaluation, bridging user intent and search outcomes at enterprise scale. Currently at Moveworks (acq. by ServiceNow), previously Signos, Nvidia, and Goldman Sachs.”
Moveworks(acq. by ServiceNow)
Senior ML Engineer – Enterprise Search Relevance
Mountain View, CA · Jan 2025 – Present
LLM AnnotationSearch RankingLLM-as-JudgeGenAI
›Implemented an LLM-based relevance annotator for a search ranking task with batched listwise labeling, yielding a 10× speedup and enabling rapid creation of large training datasets with minimal human effort.
›Reduced human labeling effort by ~60% by building a hybrid annotation workflow that uses LLM pre-annotation and escalates only ambiguous cases to human annotators for high-confidence eval sets.
›Built a summary-quality evaluation pipeline for search-result summaries using LLM-as-judge and reference-less metrics (summary relevance, groundedness, citation validity), improving summary relevance from 60% → 90%.
›Shipped a user-facing keyword-highlighting feature for search results by extracting document entities and query tokens to surface relevant terms in the UI.
›Created a synthetic query generation pipeline to augment feature-specific queries (e.g., temporal, author) and add training signal for search ranking models.
Signos
Machine Learning Engineer
Remote, CA · Apr 2023 – Dec 2024
RAGDeep LearningAnomaly DetectionRecommendation
Nvidia
Data Scientist Intern
Santa Clara, CA · Jun 2022 – Sep 2022
Time-seriesKubeflowCausal ML
Goldman Sachs
Data Scientist, Analyst
Bengaluru, India · Jun 2018 – Aug 2021
BERTNLPBig Data EngineeringPySpark
Samsung R&DInstitute, Noida
Summer ML Intern
Noida, India · May 2017 – Jul 2017
CNNTensorFlowOn-device ML
IIST
Summer Research Intern
Indian Institute of Space Science & Technology
Thiruvananthapuram, India · May 2016 – Jul 2016
Machine LearningHPCEnsemble Methods
Clickbait Spoiler GenerationSep – Dec 2022
SemEval'23 Shared Task
2-stage pipeline using text classification, QA, and ranking models. Fine-tuned DeBERTa, RoBERTa, DistilBERT — achieved 36.4 BLEU-4, surpassing benchmark.
Neural Collaborative FilteringJan – Mar 2022
Recommendation Systems
Implemented NCF from He et al. 2017 — GMF, MLP, and NMF models evaluated on 1M MovieLens dataset with comparative analysis.
NER with BiLSTM-CRFMar – May 2022
CoNLL-2003 Shared Task
Custom BiLSTM-CRF for NER with forward algorithm for partition function and Viterbi decoding for inference.
Sequence Tagging with HMMMar – May 2022
Statistical NLP
Tri-gram HMM with Viterbi decoding and smoothing (Laplace, Katz Back-off, Linear Interpolation). Context-aware N-gram LM with OOD analysis.
Adverse Food Events AnalysisSep – Dec 2021
FDA Data · EDA
EDA of FDA Adverse Food Events data (2004–2020). Identified serious outcome patterns across product categories and age groups.
Brawlhalla Elo TrackerFeb – Apr 2021
Full-stack Web App
Full-stack MERN app using Brawlhalla API to track player ratings with interactive leaderboard. Deployed on Netlify & Heroku.
GenAI & LLMsLarge Language ModelsLLM Fine-tuningLLM-as-JudgePrompt EngineeringRAG PipelinesLLM EvaluationSynthetic Data Generation
ML & SearchInformation RetrievalSearch & RankingLearning to RankRecommendation SystemsAnomaly DetectionDeep Learning
NLPNamed Entity RecognitionText ClassificationTransformersHugging FaceBERT / RoBERTaSeq2Seq
FrameworksPyTorchscikit-learnLangChainKubeflowMLflowW&BDatabricks
Data & CloudPySparkSparkSQLKafkaAWSGCPSQL
UC San Diego
MS · Electrical & Computer Engineering
Major: Machine Learning & Data Science
Sep 2021 – Mar 2023
3.91 / 4.0
GPA
IIT (BHU) Varanasi
B.Tech · Electronics Engineering
Jul 2014 – May 2018
8.81 / 10
GPA
rajasvi.xyzRajasvi Vinayak Sharma · 2026