THE PRACTICE

Engineering judgment for AI-assisted teams.

Tiercel Labs helps engineering teams understand and improve what AI coding tools change: delivery speed, code quality, review effort, and cost. Raffay Sajjad leads the assessment and the implementation.

✳ WHY THIS WORK

Generating code is one part of delivering software. Someone still has to understand the change, review it, test it, and maintain the system it joins. That is where we focus.

A faster first draft can coexist with larger PRs, more rework, or higher tool spend. Whether that tradeoff is worthwhile depends on the team, the task, and the evidence. We help engineering leaders examine those tradeoffs in their own workflow.

The starting point is a small, fixed-scope assessment. We establish a baseline, identify the friction, and implement one agreed improvement. Further work follows from the findings.

OUR FOCUS

Velocity. Quality. Cost. Together.

Optimizing one number can hide a problem elsewhere. We examine the complete path from a task to accepted, maintainable code.

Delivery effectiveness

Task-to-merge time, reviewer effort, retries, and handoffs. We look for where AI-assisted work saves time and where it moves effort to someone else.

Codebase health

Duplication, abstraction reuse, architecture drift, code churn, and CI failures. Findings come from reviewing changes in the context of the repository.

Workflow economics

Tool spend, model choice, context overhead, and human review. We weigh the cost of a usable result against quality and operating constraints.

HOW WE WORK

The engineer you brief does the work.

A founder-led practice with a bounded first engagement, direct communication, and findings your team can inspect.

One owner, end to end

Raffay runs the assessment, reviews the code and workflow, and implements the agreed change. Technical questions stay with the person doing the work.

Evidence before a recommendation

We use available usage data, representative changes, and developer interviews. Every finding distinguishes observations, estimates, and unanswered questions.

A small first scope

Begin with one team and one repository. Agree the evaluation questions, access, deliverables, and fixed fee up front. Broader implementation is a separate decision.

A usable handover

You receive the baseline, prioritized findings, agreed configuration or code changes, and a measurement runbook. Your team can repeat the checks and decide what to do next.

PRINCIPAL

Raffay Sajjad

Founder · Technical lead · Hands-on engineer

Raffay brings technical leadership at Trafilea across architecture, delivery, experiments, and production systems. That background informs how he examines changes across a team and a codebase.

He built Finly AI end to end, spanning mobile applications, backend services, infrastructure, and analytics. Building and operating a product gives the assessment a practical grounding in the consequences of engineering decisions.

His current focus is agentic development: context and repository instructions, generated-code quality, model selection, local models, and the economics of retries and review. Tiercel brings that work into a focused assessment for engineering teams.

These are founder credentials and areas of practice. They are not claims of client audit outcomes or guaranteed savings.

Focus: AI engineering effectiveness: velocity, quality, and cost
Background: Technical leadership at Trafilea; built Finly AI
Delivery: Founder-led assessment and implementation
First step: AI Engineering Efficiency Audit
Mode: Remote delivery through a US legal entity

WORKING PRINCIPLES

Measure what the team actually has to do.

Accepted work matters

Generated lines and accepted suggestions are activity signals. Review, rework, and maintainability belong in the assessment of delivery.

Cost needs context

A cheaper model is only useful if it meets the task's requirements. We include retries, review effort, and local infrastructure where relevant.

Attribution has limits

PR history alone cannot reliably identify AI-written code or show that AI caused a performance change. We document those limits and avoid false precision.

Access should be proportionate

We agree what evidence is needed before requesting access. Redacted exports and guided walkthroughs are alternatives when direct repository access is unsuitable.

Changes need an evaluation

An instruction update or model switch should be checked on representative tasks. The handover includes a plan to evaluate its effect after the assessment.

The team keeps the work

Findings, agreed code or configuration, and runbooks stay with you. Continuing with Tiercel is an option, not a dependency built into the delivery.

WHO THIS IS FOR

Engineering leaders with AI already in the workflow.

The best fit is an established software team using coding assistants or agents, with a concrete question about delivery, code quality, or spend and an internal owner for the work.

CTOs & VPs of Engineering

You need an evidence-based view of AI tooling investment, its tradeoffs, and where the next improvement is worth making.

Engineering directors

You see larger changes, review bottlenecks, or inconsistent agent output and want to understand what is happening in the team's day-to-day work.

Platform & developer-experience leads

You own shared tooling, repository instructions, model access, or CI checks and need a focused assessment of how those pieces work together.

Team size is less important than a real workflow, usable evidence, and someone who can act on the findings. This offer is focused on engineering teams already using AI coding tools.

THE FIRST ENGAGEMENT

AI Engineering Efficiency Audit.

Agree the questions

Choose one team, one repository, and the decisions the assessment should inform. Confirm available data, minimum access, the fixed fee, and the timeline.

Build the baseline

Review representative PRs, tool or usage exports, CI history, agent configuration, and team interviews. Record evidence gaps alongside the measurements.

Assess and improve

Prioritize findings by likely impact, effort, and confidence. Implement one bounded workflow or configuration change agreed in the scope.

Leave a measurement plan

Walk through the findings, hand over the changes and runbook, and define how to evaluate the result. Scope any larger implementation separately.

AT A GLANCE

A focused engineering practice.

Engagement

Fixed-scope assessment, typically one to two weeks after access is ready

Delivery

Raffay owns the assessment, agreed implementation, and handover

Scope

AI-assisted delivery, code quality, agent configuration, and workflow economics

Outcome

A baseline, prioritized findings, one bounded improvement, and a repeatable evaluation

COMPANY

Registered details.

Tiercel Labs is operated by RAFFAY LLC. Engagements are delivered remotely.

Legal name: RAFFAY LLC
Operating as: Tiercel Labs
Jurisdiction: Wyoming, United States
Registered address: 30 N Gould St Ste R, Sheridan, WY 82801
Phone: (323) 739-1571
Email: contact@tiercel.io
D-U-N-S: 13-969-3583

START HERE

Understand the tradeoffs.
Improve the workflow.

Tell us which coding tools your team uses and what you need to understand about cost, review, or delivery. We will help scope a useful first assessment.