100K+Deals processed daily by a multi-agent LLM system
~12%Lift in sales-rep engagement from causal recommendations
1 → 5Grew the LLM engineering function from scratch
$200M+Annual losses projected prevented for a U.S. health insurer
Senior Data Scientist with 6+ years applying scientific methods to high-stakes business problems, building production-grade ML/DL and end-to-end LLM/Agentic AI systems at scale. Hands-on across time-series forecasting, causal inference, recommendation, and large-scale data, with a product-oriented mindset and a habit of turning research into real-world impact.
Experience
May '25 — PresentAviso AI
Sr. Data Scientist
- Architected a multi-agent LLM system processing 100K+ deals daily across 15 enterprise clients, replacing rule-based logic with insight generation from CRM and engagement data, governed by evaluation and safety guardrails.
- Replaced correlation-based explainability with a causal-inference backbone (Double Machine Learning, Causal Forests) to surface actionable recommendations — driving a ~12% increase in sales-rep engagement.
- Built and scaled the LLM engineering function: defined the technical hiring bar and grew the team from 1 to 5 engineers for end-to-end development and production deployment.
Aug '22 — Apr '25Bosch Global Software Technologies
Sr. Computer Vision Engineer
- Published a research paper and filed a patent on enhancing foundation models (EVA-02), leveraging 90%+ unlabeled Bosch data to boost performance by ~5%; collaborated with Bosch Romania to further improve EVA by ~4.31% via cluster analysis and hypothesis testing on DINOv2 embeddings.
- Built a Vision Assignment Marketplace, training SWin, DeiT, and DeTr with ~6.72% average performance deviation; designed a ConvNeXt segmentation pipeline on fisheye images (~57% F1).
- Deployed an object detector and motion analyzer on a Jetson Nano edge test-bed at ~30 FPS; refactored legacy code for newer vision-transformer architectures, improving adaptability/efficiency by 14%.
Oct '20 — Jul '22HiLabs Inc.
Data Scientist II
- Engineered a multi-million-dollar data pipeline moving anomalies from large-scale SQL/RDBMS sources to HBase and Solr for ticket-based remediation, improving Data Quality Index from ~78 to ~86; projected to prevent $200M+ in annual losses for a leading U.S. health insurer.
- Deployed an enterprise-grade OCR solution (NLP + custom algorithms) on contracts after GAN-based image enhancement, generating $100K+ in business value.
Jan '20 — Mar '20GMO ResearchJapan · internship
Machine Learning Engineer Intern
- Forecast 6-hour survey response likelihood with time-aware Random Forests, Decision Trees, and DNNs; addressed class imbalance via SMOTE and custom metrics, achieving 91% efficacy on responsive marketing panels.
What I work with
Languages & libraries
PythonSQLScalaPandasNumPyScikit-Learn
Big data & MLOps
SparkHadoopHBaseSolrAWSDockerGitCI/CDGPU/HPC
ML / DL / LLM
PyTorchTensorFlowTransformersLangChainLangGraphRAGAgentic AIMCP
Models & paradigms
Time-Series ForecastingRecommendationCausal InferenceAnomaly DetectionComputer Vision
Publications & patents
Patent · co-authoredDimensionality-reduction techniques to detect and localize failure regions in segmentation outputs from foundation models, improving interpretability and reliability.
ResearchA two-stage fine-tuning approach improving foundation-model segmentation performance by 5%+ mIoU using unlabeled data — semi-supervised learning.
Education
B.Tech, Mechanical EngineeringIndian Institute of Technology, Mandi · Minor in Intelligent SystemsAug '16 — Jul '20
References available on request. Email is the fastest way to reach me.