Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I’m Akarsh Kollana, a Senior AI/ML Engineer in the US with 9+ years of experience building production-grade Generative AI solutions. My focus is on RAG architectures, agentic/multi-agent workflows, and enterprise ML platforms, with a strong emphasis on latency, cost optimization, and reliable operations. Across retail/e-commerce, supply chain, and healthcare, I’ve led end-to-end AI/ML delivery—from feature pipelines and model training to deployment, monitoring, and evaluation/guardrails. I also work extensively with document intelligence and structured data extraction, and I enjoy bridging engineering and domain needs to create practical, grounded AI systems for real teams.…I’m Akarsh Kollana, a Senior AI/ML Engineer in the US with 9+ years of experience building production-grade Generative AI solutions. My focus is on RAG architectures, agentic/multi-agent workflows, and enterprise ML platforms, with a strong emphasis on latency, cost optimization, and reliable operations. Across retail/e-commerce, supply chain, and healthcare, I’ve led end-to-end AI/ML delivery—from feature pipelines and model training to deployment, monitoring, and evaluation/guardrails. I also work extensively with document intelligence and structured data extraction, and I enjoy bridging engineering and domain needs to create practical, grounded AI systems for real teams.

Akarsh Kollana

Data Scientist, AI Engineer, AI Developer, +7





I’m Akarsh Kollana, a Senior AI/ML Engineer in the US with 9+ years of experience building production-grade Generative AI solutions. My focus is on RAG architectures, agentic/multi-agent workflows, and enterprise ML platforms, with a strong emphasis on latency, cost optimization, and reliable operations. Across retail/e-commerce, supply chain, and healthcare, I’ve led end-to-end AI/ML delivery—from feature pipelines and model training to deployment, monitoring, and evaluation/guardrails. I also work extensively with document intelligence and structured data extraction, and I enjoy bridging engineering and domain needs to create practical, grounded AI systems for real teams.…I’m Akarsh Kollana, a Senior AI/ML Engineer in the US with 9+ years of experience building production-grade Generative AI solutions. My focus is on RAG architectures, agentic/multi-agent workflows, and enterprise ML platforms, with a strong emphasis on latency, cost optimization, and reliable operations. Across retail/e-commerce, supply chain, and healthcare, I’ve led end-to-end AI/ML delivery—from feature pipelines and model training to deployment, monitoring, and evaluation/guardrails. I also work extensively with document intelligence and structured data extraction, and I enjoy bridging engineering and domain needs to create practical, grounded AI systems for real teams.

Available to hire

I’m Akarsh Kollana, a Senior AI/ML Engineer in the US with 9+ years of experience building production-grade Generative AI solutions. My focus is on RAG architectures, agentic/multi-agent workflows, and enterprise ML platforms, with a strong emphasis on latency, cost optimization, and reliable operations.

Across retail/e-commerce, supply chain, and healthcare, I’ve led end-to-end AI/ML delivery—from feature pipelines and model training to deployment, monitoring, and evaluation/guardrails. I also work extensively with document intelligence and structured data extraction, and I enjoy bridging engineering and domain needs to create practical, grounded AI systems for real teams.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Language

English

Fluent

Work Experience

AI/ML Engineer – Generative AI at Tech Mahindra / Applied materials

June 1, 2026 - Present

Designed and deployed a RAG-based Generative AI solution using Gemma 3, Llama 3.2, Ollama, Qwen3-Embedding, EmbeddingGemma, vector search, and semantic chunking, reducing manual knowledge lookup effort ~35%. Improved retrieval relevance and response speed by testing chunk sizes, embedding models, top-k retrieval, reranking, and prompt structure, reducing average response latency ~15% while keeping outputs grounded in source content. Built an AI-powered document intelligence workflow for 50+ technical PDFs using Python, PyMuPDF, OCR, Pydantic, and structured LLM outputs to extract and validate materials information, cutting manual review effort ~30%. Developed a multi-agent GenAI workflow using Azure OpenAI, LangChain, and LangGraph to analyze supplier and network risk across 6,800+ sub-tier entities and surface risk concentrations across downstream tiers. Integrated Hugging Face MCP with LangGraph agents for controlled tool/model access and structured tool execution. Added evaluation,

Senior AIML Engineer at Zoro.com (Grainger subsidiary)

August 1, 2022 - Present

Led the development of a cloud-native AI/ML platform on AWS unifying transactional, clickstream, inventory, and vendor data; reduced reporting and feature-processing latency from hours to under 20 minutes. Implemented RAG-based AI assistants and semantic search workflows using LangChain, LangGraph, OpenAI, Amazon Bedrock, MCP, and OpenSearch vector search. Built scalable ML pipelines with Spark Structured Streaming, Airflow, and dbt to support demand forecasting, recommendations, and operational analytics. Achieved 99.9% SLA and cut infrastructure costs by optimizing Redshift and Snowflake workloads; automated model deployment and retraining with MLflow, Docker, Kubernetes, Jenkins, Terraform, and GitHub Actions. Implemented data quality and observability using Great Expectations, Soda Core, Datadog, CloudWatch, and MLflow tracking, reducing production data incidents by 45%.

Senior AI/ML Engineer at Zoro.com (Grainger subsidiary)

August 1, 2022 - Present

Built a cloud-native AI/ML platform on AWS using Python, PySpark, Databricks, Redshift, Glue, and Snowflake to unify transactional, clickstream, inventory, and vendor data, reducing reporting and feature-processing latency from hours to under 20 minutes. Developed RAG-based AI assistants and semantic search workflows using LangChain, LangGraph, OpenAI, Amazon Bedrock, MCP, and OpenSearch vector search, enabling faster knowledge retrieval and reducing manual support investigation effort by 35%. Designed scalable ML and feature engineering pipelines with Spark Structured Streaming, Airflow, and dbt to support demand forecasting, recommendation systems, customer segmentation, and operational analytics across merchandising and supply chain teams. Improved platform performance and reduced annual infrastructure costs by $320K through Redshift and Snowflake query optimization, partitioning strategies, clustering design, and workload tuning while maintaining 99.9% SLA availability. Automated m

Data Scientist at Target

January 1, 2021 - July 1, 2022

Built scalable ETL/ELT and ML feature pipelines using Python, PySpark, Snowflake, Hive, Airflow, and dbt to process clickstream, pricing, ERP, and customer behavior data, improving ingestion throughput by 40% for analytics and personalization workloads. Designed dimensional data models and reusable feature datasets in Snowflake using star schema, SCD Type 2, and CDC-based ingestion frameworks, helping Analytics and Data Science teams standardize KPIs and accelerate self-service reporting. Developed machine learning solutions with TensorFlow, PyTorch, and scikit-learn for recommendation systems, pricing optimization, customer segmentation, and conversion forecasting, contributing to a 15% increase in targeted campaign conversions. Implemented real-time streaming pipelines using Kinesis, Snowpipe, and AWS Lambda to process order and inventory events with under one-minute latency during peak retail traffic periods. Improved Snowflake performance and reduced compute costs by 30% through cl

Data Analyst/Analytical Engineer at Y-Prime

June 1, 2017 - December 1, 2020

Analyzed large-scale EHR/HL7/FHIR data to identify trends in patient outcomes, provider performance, and utilization. Built healthcare analytics pipelines on AWS and Databricks using Spark SQL, Airflow, and ETL frameworks to transform raw clinical datasets into reporting-ready data marts. Developed dimensional models and KPI reporting layers with dbt and Databricks SQL, improving accessibility of care-management analytics. Created Tableau and Power BI dashboards for provider scorecards, claims analysis, patient engagement, and utilization reporting, reducing manual reporting turnaround by 40%. Automated data validation and reconciliation using Great Expectations and SQL-based quality controls, improving reporting accuracy and reducing defects. Ensured HIPAA-compliant workflows with PHI masking, RBAC, audit logging, and secure data handling across AWS and Databricks while collaborating with clinical, compliance, and analytics teams.

Data Analyst / Analytical Engineer at Y-Prime

June 1, 2017 - December 1, 2020

Analyzed large-scale EHR, HL7, FHIR, CCD, and healthcare claims data using Python, SQL, PySpark, and Databricks to identify trends in patient outcomes, provider performance, and healthcare utilization patterns. Built healthcare analytics pipelines on AWS and Databricks using Spark SQL, Airflow, and ETL frameworks to transform raw clinical and claims datasets into reporting-ready data marts for operational and regulatory reporting. Developed dimensional data models and KPI reporting layers with dbt and Databricks SQL, improving accessibility of clinical and encounter analytics for care management and business operations teams. Created Tableau and Power BI dashboards for provider scorecards, claims analysis, patient engagement, and utilization reporting, reducing manual reporting turnaround time by 40%. Automated healthcare data validation and reconciliation checks using Great Expectations, Python, and SQL-based quality controls, improving reporting accuracy and reducing downstream data