01
Kittl Agentic AI
Current hands-on production experience: owned the Agentic AI project at Kittl, from architecture and technical direction through to production delivery.
AI-NATIVE DELIVERY SYSTEM SPRINT
Turn coding agents into a repeatable delivery system.
For small software and AI product teams that already use coding agents, but still rely on founders or senior engineers to repeatedly supply context, correct direction, rerun checks and carry tasks across the finish line.
We take one recurring production workflow and turn the repeated human guidance around it into a maintainable agent-ready system.
PRODUCTION AI ENGINEERING
Production AI engineering — not prompt demos.
02
AI-assisted technical sign-off
Designed an AI-assisted engineering sign-off workflow adopted across approximately five product teams and used on 23 projects.
03
AI Change Verification — open source
We built and open-sourced an evidence-accountable verification workflow for AI-assisted code changes before human review — with deterministic evidence, unresolved-risk handling and explicit human judgment.
AI Change Verification on GitHubWHEN THE AGENT IS FAST BUT THE HUMAN IS STILL THE SYSTEM
AI can write more code. That does not mean you have delegated the workflow.
01
Context keeps resetting
Architecture, repository rules, previous decisions and task intent still have to be reconstructed or re-explained every time the agent starts meaningful work.
02
Senior engineers still have to steer execution step by step
The agent produces quickly, but a founder or senior engineer repeatedly corrects direction, resolves ambiguity, reruns checks and carries the task to completion.
03
Good AI practices remain personal
Useful prompts, skills, conventions and verification habits live in individual setups instead of becoming a shared system the team can maintain.
If coding agents are not already materially used in real development, this is probably not the right Sprint.
THE SPRINT
Turn repeated senior guidance into a system agents can reuse.
We select one recurring workflow where human guidance is being repeated and make the missing context, constraints, checks and escalation points more explicit and reusable.
The intervention depends on the actual workflow. We do not install a generic “agent stack”.
The goal is not more AI usage. It is more engineering work delegated safely with less repeated senior intervention.
- 01
Shared repository context and skills
When agents repeatedly miss architecture, conventions, domain rules or prior decisions.
- 02
Clearer task and acceptance contracts
When too much direction is reconstructed interactively during execution.
- 03
Tool access where it removes real manual work
CLI, APIs or MCP-style connections only where they remove repeated copy-paste or context switching.
- 04
Deterministic checks and hooks
Tests, linting, type checks, structural rules or other machine-verifiable constraints before human attention is required.
- 05
Stop, escalation and human-judgment boundaries
So the agent knows what it may complete, what requires evidence and what must return to a person.
- 06
One repeatable delivery loop
For example: task → implementation → checks → evidence → human decision.
Five business days, one recurring workflow.
- 01
Select the workflow and baseline
Choose one recurring workflow and identify where founder/senior time is still repeatedly spent around the agent.
- 02
Make context and constraints inspectable
Turn the most important implicit knowledge, decisions and non-negotiables into a form the workflow can reliably use.
- 03
Build the missing checks and agent loop
Implement the smallest useful skills, tooling, checks, hooks or escalation logic required by that workflow.
- 04
Run it on live work
Test the new system on real tasks rather than a demo repository.
- 05
Handoff
Leave the working setup, evidence, unresolved limitations and a short operating runbook with the team.
FIT
Best for a small team that already depends on coding agents.
Often 2–10 people, but headcount is not the deciding factor. The real question is whether repeated senior guidance has become part of the cost of using agents.
GOOD FIT
- Coding agents are already part of day-to-day engineering.
- A founder, CTO or senior engineer repeatedly provides the same context or corrections.
- There is at least one recurring workflow worth making more autonomous.
- The team has a real production repo and is willing to test changes on real work.
- The team wants implementation, not generic AI training.
NOT A FIT
- The team is only beginning to experiment with AI coding tools.
- There is no recurring workflow to improve.
- The main problem is product-market fit rather than engineering execution.
- The request is for generic prompting training or a company-wide AI transformation programme.
WORKING DIRECTLY WITH ENGINEERS
The engineers doing the work
AI & Software Engineering Lead
Yaroslava Suspitsina
Senior software engineer focused on production AI systems, agentic architecture and engineering workflows. She currently works at Kittl, where she owned the Agentic AI project from architecture and technical direction through to production delivery. She has also designed AI-assisted review, testing and quality workflows used across multiple product teams.
- Current Kittl Agentic AI project ownership and production delivery.
- AI-assisted technical sign-off used across approximately five product teams and 23 projects.
- Agentic architecture, testing, CI quality automation, Playwright, debugging and engineering enablement.
AI/ML Engineer
Valerii Safonov
AI/ML engineer with experience across machine learning, quantitative modelling, NLP/RAG, computer vision and production-oriented AI systems. MSc in Computer Science with a machine-learning focus, with applied AI engineering experience dating back to at least 2020.
- MSc in Computer Science with a machine-learning focus.
- Applied AI/ML work across quantitative modelling, NLP/RAG, computer vision and production engineering.
- Earlier applied technology-trend detection and evaluation work spanning data, architecture and engineering methods.
No handoff from sales to a junior delivery team. The people shown here work on the engagement.
FIXED-SCOPE SPRINT
One production repo. One recurring workflow. Five business days.
€2,500 fixed
+ VAT where applicable
5 business days · 1 production repo · 1 recurring workflow · implementation in the real environment · live trial · handoff.
If we cannot first identify a recurring workflow where this level of intervention could reasonably be useful, this Sprint is not the right fit.
No ongoing retainer required. Additional work is scoped separately only if there is a clear reason to continue.
Questions before the Sprint
Is this AI training?
01No. The Sprint assumes your team already uses coding agents. We work on the engineering system around one real workflow rather than teaching generic prompting.
Which coding agents or tools do you support?
02The Sprint is tool-neutral. The exact intervention depends on the tools already used by the team and may involve repository context, agent skills, CLI/API integrations, MCP-style connections, checks, hooks or workflow changes where useful.
Is this only for 2–10-person teams?
03That is a common fit, not a hard rule. The deciding factor is whether repeated founder or senior-engineer guidance has become part of the cost of using agents.
Do we need to migrate to a new platform?
04No. We prefer improving the existing engineering environment unless a specific tool change is clearly justified by the workflow.
What access do you need?
05Enough scoped access to understand and improve the selected repository workflow. Existing security controls and human approval boundaries remain in place.
Do you guarantee a productivity increase?
06No fixed percentage is promised. The Sprint is deliberately bounded so the team can test whether one concrete system change reduces repeated human effort in the selected workflow.
What if our bottleneck is broader than agent babysitting?
07The AI Engineering Workflow Sprint is designed for broader delivery bottlenecks involving context, requirements, architecture, rework, verification, review or coordination.
What happens after five business days?
08The Sprint ends with handoff. The team keeps the implemented setup, evidence and runbook. There is no automatic consulting commitment.
Contact
Show us the agent workflow that still needs too much human guidance.
In a few sentences, tell us your team size, the AI/coding tools you already use, and one recurring workflow where context, corrections, checks or human intervention keep repeating.