AI Product Leader | 0→1 Innovation | Agentic Systems & LLM Evaluation | Building Useful, Trustworthy Products
I am an AI Product Leader who turns complex problems into clear product direction, practical workflows, and measurable outcomes. My experience spans software engineering, technical leadership, and product management across enterprise SaaS, cybersecurity, analytics, banking & financial services, and industrial automation domain.
Here, I build and share AI applications, agentic workflows, and evaluation systems. I am interested in end-to-end customer journey: understanding the problem, forming hypotheses, exploring potential solutions, prototyping, deciding what to build, validating how it behaves, and learning where it falls short.
Each project is a place to share my product thinking, the decisions behind the build, and what I learn along the way.
- AI agents that support real decisions and everyday workflows
- LLM evaluation, error analysis, and human feedback loops
- Spec-driven development that connects user needs to implementation
Product & Leadership Product strategy · Customer discovery · Roadmaps · Prioritization · Success metrics · Cross-functional leadership · Team development
AI & Evaluation LLM evaluation · Evaluation harnesses · Error analysis · Prompt comparison · Regression detection · Human-in-the-loop workflows · MCP · LangChain
Engineering & Delivery Python · TypeScript · SQL · Java · JavaScript · C# · GitHub Actions · CI/CD · Azure
Product Collaboration Jira · Confluence · Miro · Figma · Agile / SAFe
Decision support for releasing AI prompt changes
Helps teams assess whether a prompt change improves AI behavior and where it introduces regressions, providing evidence for release decisions.
My product focus: defining what “better” means, weighing improvements against regressions, and making tradeoffs visible to decision-makers.
Explore the project → https://github.com/aastha0208/prompt-eval-gate
A product case study in AI-assisted customer support for evaluating and improving AI agents
Explores how an AI agent can help shoppers with orders, returns, and refunds while navigating store policies and knowing when to escalate to a person.
As part of the AI Evals for Engineers & PMs course, I analyzed customer conversations, identified failure patterns, and refined the agent’s instructions.
My product focus: balancing convenience with customer trust, defining boundaries for autonomous actions, and using observed failures to improve the support experience.
Explore the project → https://github.com/aastha0208/cartwheel-support-agent
AI decision support for choosing the right automation layer
Helps engineering teams decide what to automate and at which layer, using ticket requirements and engineering guidelines to recommend an approach with a clear rationale.
My product focus: reducing decision overhead, fitting into existing workflows, and keeping engineers in control of the final choice.
Explore the project → https://github.com/aastha0208/automation-layer-adviser
One trusted view of release health for go/no-go decisions
Gave release stakeholders, from product managers and owners up to the VP of Engineering and Product and the Release Review Committee, a one-click view of each monthly release, with drill-down to the failures and blockers behind it.
My product focus: aligning teams and leadership on what "pass rate" and "complete" mean, so certification-week decisions rested on one agreed set of numbers instead of competing ones.
Explore the project → https://github.com/aastha0208/release-health-metrics
Hands-on exercises I use to stay close to the technology I make product decisions about:
- LangChain Agents Lab: agents, tools, memory and model-routing middleware, with notes on where each fits in a product
- MCP Research Assistant Lab: an MCP server and client, and what the protocol means for governing what AI can reach
- Programming Concepts Across Languages: 12 core concepts in five languages, every example tested automatically