LLM Specialist & ML Engineer at Tiger Analytics (2023-05 – 2026-04)
Collaborated closely with cross-functional teams—including product managers, domain experts, and business stakeholders—to translate complex ML research into practical, high-impact AI features that align with user needs and strategic product goals.
- Implemented RAG pipeline for a Crypto Trading AI Multi-Agent and integrated real-time data pipeline and building an ensemble forecasting system with LSTM, Prophet, and ARIMA, along with ensemble model for risk assessment.
- Engineered a end to end RAG-MCP pipeline for a DeFi Risk Agent, enabling real-time estimation of Ethereum wallet risk by integrating transaction history, balances, and multi-protocol active positions.
- Designed and developed a robust, scalable MLOps pipeline-from data ingestion to model deployment-using MLflow and Kubeflow on an AWS EKS cluster.
- Spearheaded the development of multiple Generative AI applications using diverse tech stacks, including LLMOps tools, Vector Databases, and both proprietary and open-source LLMs, to enable capabilities such as entity extraction, question answering, data analysis, and intent detection.
- Designed and deployed a robust RAG-based system for a multi-agent assistant in healthcare settings, integrating external tools and services via Model Context Protocol (MCP) to enable dynamic, context-aware decision-making.
- Built a virtual environment using OpenAI Gym for reinforcement learning of a trading agent and implemented Direct Preference Optimization (DPO) to enhance the accuracy of large language models (LLMs).
- Fine-tuned an open-source LLM using QLoRa with transformer and improve the retrieval accuracy using hybrid search in RAG pipeline.
- Built an ETL data pipeline to ingest data from multiple DeFi protocols using ApacheAirflow, Kafka, and Spark, and efficiently stored it in InfluxDB for real-time analysis.
- Contributed in the development and deployment of machine learning models, fine-tuning pretrained LLMs on domain-specific datasets onAWS SageMaker to improve performance.
Data & AI Engineer at Mercari (2021-07 – 2023-05)
- Implemented MLOps pipelines with Kubeflow, MLflow, and Seldon Core on EKS, automating model training, deployment, and monitoring for scalable and reliable production workflows.
- Designed and deployed a hybrid item-recommendation engine combining Elasticsearch BM25, vector-based semantic search, and reranking models, powered by a real-time PySpark–Kafka–Snowflake pipeline on AWS.
- Optimized retrieval relevance, latency, and index freshness through ANN tuning, fusion scoring, and dual-index synchronization, increasing customer engagement by 9% and boosting sales by 12%.
- Fine-tuned CLIP (Contrastive Language-Image Pre-Training) with domain specific dataset from e-commerce platform to significantly improve the embedding calculation accuracy.
- Engineered end-to-end ETL pipelines with PySpark to transform and enrich raw data into high-value features for recommendation and predictive modeling.
- Collaborated with cross-functional teams to preprocess data, fine-tune ML models, and deploy production-ready models, including TensorFlow Lite solutions for client-side inference.
- Reduced fraudulent activity by 9% by designing and deploying a real-time anomaly detection system using traditional Machine Learning Algorithms for financial transactions, significantly improving security and asset protection.
AI & Web3 Engineer at Token Metrics (2019-04 – 2021-07)
- Designed and developed an automated ETL pipeline using Snowflake, Apache Spark, and Kafka Streams APIs on AWS EC2 to ingest and process large-scale datasets from DeFi protocols such as Uniswap V3, Aave, Liquity, and Lido.
- Implemented comprehensive data quality validation and monitoring systems to ensure accuracy, consistency, and reliability across all inputs for forecasting models.
- Deployed forecasting models via AWS SageMaker with canary deployment strategies and CloudWatch integration, reducing rollout risks and maintaining continuous uptime.
- Engineered a scalable, containerized microservices architecture with Docker and Kubernetes, accelerating deployment cycles and improving system resilience.
- Developed a real-time crypto market intelligence dashboard powered by The Graph (Subgraph), delivering transparent on-chain analytics and actionable market insights.
Full Stack Developer at Cloudera (2017-05 – 2019-03)
- Designed and developed new features for a scalable, high-performance E-commerce Saas platform using Node.js and Python, collaborating with cross-functional teams to deliver solutions aligned with strategic business goals.
- Contributed to the successful migration from a monolithic to microservices architecture, leveraging Apache Kafka for inter-service communication to enhance system scalability, modularity, and maintainability.
- Implemented modern CI/CD pipelines using Docker, GitHub Actions, and Argo CD for automated deployments to EKS, enabling faster and more reliable release cycles.
- Managed infrastructure-as-code using Terraform, streamlining AWS resource provisioning and reducing deployment complexity and operational overhead.
- Integrated multiple global payment providers (Stripe, Adyen, PayPal), ensuring full PSD2 and GDPR compliance, while enabling secure, multi-currency transactions across international markets.
