Prakash Kantumutchu

Training and mentoring

Practical sessions drawn from systems I build and run. I taught MLOps professionally before my current role, and I train engineers on AI tooling at work.

Formats: one-to-one mentoring, group workshop, training for an organisation. Training builds skills; it does not guarantee a job, an interview or a pay rise.

AI engineering for working software engineers

For: Backend and full-stack engineers who can ship a web service and want to add LLM features properly.

Before you start: Comfortable with one backend language and HTTP APIs. No machine-learning background needed.

You'll be able to

  • Call models through a provider-agnostic boundary with timeouts, retries and fallbacks
  • Validate model output against a schema and decide what happens when it fails
  • Keep personal data out of places it should not go
  • Build a small evaluation set and use it before every prompt change

Example sessions

  1. How LLM APIs fail in production
  2. Structured output and validation
  3. Privacy boundaries and tracing
  4. Evaluation sets and release gates
  5. Cost and latency budgets

Building LLM features that fail safely

For: Teams about to put an LLM feature in front of real users.

Before you start: A team that already ships software to production.

You'll be able to

  • Separate what the model decides from what the system records
  • Design fallbacks for model errors, timeouts and outages
  • Give agents read-only, bounded access to data

Example sessions

  1. Deterministic-first design
  2. Fail-closed checks
  3. Tool and data access for agents
  4. Shadow-mode rollouts

A technical portfolio that survives a screen

For: Engineers moving into software or AI roles from another field, as I did.

Before you start: At least one project you have built or contributed to.

You'll be able to

  • Choose the two or three projects that actually show your capability
  • Write a case study that separates your work from the team’s
  • Clean up a GitHub profile so it supports, rather than contradicts, your CV

Example sessions

  1. Evidence over adjectives
  2. Writing case studies
  3. GitHub and LinkedIn alignment

Send a training enquiry

Fields marked optional can be left blank.

Is this for you or for an organisation?

Please do not include ID numbers or confidential employer material.