Applied AI · Product Strategy · Technical Leadership

Turning ambiguous AI ideas into systems that actually work.

I’m Ashwin Parthasarathy — a technical product leader with 20+ years across Amazon, Google, Microsoft, and technology consulting, focused on applied AI systems — from model behavior and prompting to agent design and production decision systems.

Explore My Work Read My AI Notes Connect

Recent impact

Selected proof points from applied AI, developer productivity, data systems, and product execution.

$14M+ AI/ML business impact

Led development of a machine learning model for product end-of-life prediction, creating $1M in annual selection-monitoring savings and enabling $13M in incremental worldwide PCOGS.

Agentic AI for sourcing traceability

Led development of an AI agent that helped retail teams understand the rationale behind sourcing decisions and reduced KTLO overhead by 50% for the Sourcing Optimization team.

Developer productivity at scale

Led and built tools, metrics, and data pipelines across Google and Microsoft, including debugging systems, adoption metrics, and enterprise-scale AI-driven product search.

Where I create leverage

I focus on problems where product judgment, technical depth, and execution discipline all have to come together — especially in AI systems where the hardest problems are often product, data, workflow, and adoption problems.

AI Product Strategy

Identifying high-value AI opportunities, clarifying the customer problem, and shaping roadmaps that connect technical possibility to real business outcomes.

ML Systems Thinking

Looking beyond model scores into data quality, evaluation, failure modes, human workflows, and the operational reality of deploying machine learning.

Execution Leadership

Driving cross-functional work across engineering, science, business, and operations teams with clear ownership, crisp trade-offs, and measurable outcomes.

Professional narrative

My career has moved from building systems, to leading products, to applying AI in ways that are practical enough to ship and valuable enough to matter.

Amazon

Applied AI for retail decision systems

Currently leading technical product work on ML and agentic AI systems, including product lifecycle prediction and sourcing traceability for retail teams.

Google

Developer productivity, metrics, and platform adoption

Built adoption metrics, optimized data pipelines, and led product work for developer tooling, debugging, framework convergence, and large-scale engineering productivity.

Microsoft

AI-driven commerce products

Led end-to-end development of Bing for Commerce’s AI-driven Product Search, from concept and executive sponsorship through pilot launch and customer integration.

Earlier

Architecture, consulting, and emerging technology

Led architecture and technology strategy across IoT, automotive, smart assistants, precision agriculture, emergency response systems, and enterprise cloud solutions.

Selected projects

Work that shows how I think: frame the problem clearly, understand the system around it, and build toward measurable outcomes.

Flagship case study

Applied AI for product lifecycle and sourcing decisions

Led technical product work on ML and agentic AI systems for retail decision-making: predicting product end-of-life, improving selection monitoring, and helping teams understand the rationale behind sourcing decisions.

This is the type of AI work I’m most interested in: not a demo, not a model in isolation, but a decision system that has to connect data quality, model behavior, business workflow, explainability, and operational adoption.

Applied AI Agentic AI ML Product Strategy Retail Systems
$1M Annual savings in selection monitoring
$13M Incremental worldwide PCOGS enabled
50% KTLO overhead reduction for sourcing optimization
AI + Product Model behavior connected to real decision workflows

Machine Learning Lab

A hands-on learning lab where I build and document ML projects across model evaluation, computer vision, transfer learning, and the practical habits needed to understand model behavior.

PythonScikit-learnPyTorch

AI Blog & Notes System

A public knowledge base where I turn hands-on learning into story-driven explanations of machine learning, AI systems, evaluation trade-offs, and technical product judgment.

WritingObsidianQuartz

Robotics & FLL Tools

Experiments in robot navigation, test harnesses, grid-based movement, and engineering practices for FIRST LEGO League robot programming and team learning.

PythonPybricksRobotics

Writing

I write to break down how AI systems actually work in practice — from model behavior and evaluation to prompting, agents, and production decision systems.

Latest focus: machine learning from the inside out

My current writing series walks through machine learning concepts through actual moments of confusion, debugging, and discovery — from misleading accuracy metrics to cross-validation, regularization, model selection, and production thinking.

Visit notes.ashwinlabs.com

AI Systems & Prompting

A focused series on how large language models actually behave in practice — from prompt design to agent systems. This work reflects how I think about AI: not as isolated models, but as decision systems that need to be designed, constrained, and made reliable.

Prompt Engineering Is Not About Prompts

Why prompt engineering is really about reducing ambiguity and controlling model behavior, not asking better questions.

Read post

Why Prompts Fail (And How to Debug Them)

A practical framework to diagnose ambiguity, scope issues, and format drift — and systematically improve outputs.

Read post

Inside the ReAct Pattern

How Think → Act → Observe turns a model into an agent, and why this pattern is foundational to tool-using systems.

Read post

How Tool Calling Actually Works

What really happens when a model “calls an API” — and why separating decision from execution is critical.

Read post

From Prompts to Agents

Designing multi-step decision loops with tools, state, and control — where prompting becomes system design.

Read post

Let’s connect

I work on applied AI, ML product strategy, developer productivity, and technical leadership problems where ideas need to become real systems.

LinkedIn Read my notes GitHub