One-to-one mentorship
AI-native development & agentic engineering
Use coding agents with clear specifications and independent verification, or engineer a tool-using workflow with explicit state, permissions, and recovery.
Book a free discovery call 10 minutes · English or Bangla · No commitment
Who this is for
Working developers and teams adopting AI-assisted development, or building agentic capabilities into a product. We choose the development-workflow or agent-system focus around your goal.
Practical outcomes
We choose the outcomes around your starting point and the scope of the engagement.
- Turn a product requirement into acceptance criteria and a bounded implementation task.
- Give a coding agent the relevant repository context and review its assumptions and changes.
- Verify generated work against the requirement and investigate a plausible but incorrect result.
- Design a small agent workflow with narrow tools, trusted authorization, and durable state.
- Exercise retries, missing information, and interrupted actions without assuming that a model response proves success.
Choose your focus
- AI-native development: specifications, repository context, bounded tasks, and diff review
- Verification: acceptance criteria, meaningful tests, browser behavior, and release considerations
- Agentic engineering: tool contracts, authorization, workflow state, and execution limits
- Evaluation: representative tasks, failure cases, uncertain outcomes, and recovery
How I help you get there
Work in the real development context
I help you inspect the repository, define the behavior, and prepare a task an agent can work on without guessing about the surrounding system.
Review the engineering behind the output
We inspect changes together, challenge assumptions, trace failure paths, and choose checks that test the requirement independently of the generated implementation.
Engineer and evaluate a workflow
For an agent-system engagement, we define tools and state, connect a small working path, and examine ordinary tasks and failure cases before expanding its scope.
Example session
From a feature brief to a verified change
Use a small feature in your own project to practice an AI-native development loop. An agent-workflow session can instead focus on one bounded tool-using task.
- Define the user outcome, constraints, acceptance criteria, and relevant repository paths.
- Guide a bounded implementation, inspect the diff, and challenge unclear design choices.
- Verify the behavior and an important failure case, then record what remains unresolved.
Check your understanding
You explain what the agent was given, which assumptions you checked, and what evidence supports the result. For a workflow, you can also explain its tool boundaries and failure states.
Working outputs & progress
Depending on the agreed focus, useful outputs can include:
- A reusable task brief and a repository-context checklist.
- A reviewed implementation or a scoped workflow prototype.
- Acceptance checks or evaluation cases for the agreed behavior.
How we assess progress
Progress means you can define a useful task, recognize an unsupported claim, inspect the resulting behavior, and choose the next verification step. For agent systems, it also means understanding what the application must enforce outside the model.
Experience behind the guidance
Explore the engineering work and reasoning I bring to these discussions.
Agree on the work before we begin
Sessions are available in English or Bangla, on demand or by the hour. After the free discovery call, we agree on preparation, session scope, useful outputs, timing, and fees. Any review or support between sessions is agreed explicitly.
Learning with your team?
The same topics can become a focused workshop or a series of mentoring sessions for your engineering team.