Search relevance · Architect-led

Search relevance, retrieval and ranking consulting.

Your search box is a product decision, not a plugin. We help teams build search, retrieval and ranking systems that understand what users mean, respond fast and can be measured. Led by Manmeet Singh, who built search and recommendation systems at Apple and Instacart.

AIshar Labs helps teams improve search, retrieval and ranking in their products, including query understanding, retrieval, ranking, latency and evaluation. The work is led by Manmeet Singh, founder of AIshar Labs, whose earlier roles at Apple and Instacart centered on search, autocomplete, recommendations and personalization. Engagements run as a two-week Architecture Sprint, an Embedded Build or a Targeted Build.

Search is a product, not a feature.

Five areas of a search system, end to end.

Search is more than ranking. We look at how the system understands the question, finds candidates, orders them, stays fast and proves it works.

Area 01

Query understanding

What the user means, not only what they typed: spelling correction, intent, entities and context. From years of building search at scale, we treat this as the hardest part, and the one that most often makes AI products feel "dumb" when it is skipped.

Area 02

Retrieval

Keyword, semantic and hybrid retrieval, and how each fits your data and your latency budget. We have worked with Elasticsearch, Lucene, two-tower retrieval models and RAG pipelines, in prior roles and in consulting work. This includes the retrieval layer behind RAG systems, where an answer can only be as good as what gets retrieved.

Area 03

Ranking

Ordering results by what matters to your users and your business, including personalization where your data supports it.

Area 04

Latency

Response time as a design input, not an afterthought. As-you-type search and real-time recommendations limit which models and architectures are practical, so the latency budget comes first.

Area 05

Evaluation

Offline metrics, online experiments and human review, chosen to match what your users actually value instead of a default accuracy number. If you can't measure relevance, you can't improve it.

Delivered through the engagement models you already know.

Search relevance is an area of expertise, not a separate product, so there is no separate price list. The same models apply as on How We Work.

Where the experience comes from.

Retrieval for AI answers is covered under RAG consulting. This background comes from Manmeet's earlier roles. Apple and Instacart are not clients of AIshar Labs and do not endorse it. Read more about Manmeet or the five search relevance lessons that apply to every AI product.

An online clothing retailer's recommendations were basic or wrong. We built a system where shoppers type or speak, a search agent or voice agent calls RAG over the retailer's own data, Gemini models handle classification and decisions, and OpenAI models serve as the frontier models, all on GCP with Terraform. Click-through rate improved by 15 to 20 percent and add-to-cart by 5 to 7 percent.

Read the full story →

Questions teams ask before they book.

What is search relevance consulting?
Helping a team improve how well its search, retrieval and ranking systems match results to what users want. That covers how queries are understood, how results are retrieved and ordered, how fast they return, and how quality is measured.
Do you work on RAG and AI assistants as well as classic search?
The same questions apply: what the user means, what gets retrieved, and how you measure quality. We look at the retrieval layer that feeds the answer, because a generated answer can only be as good as what was retrieved.
How does an engagement start?
With a 30-minute conversation. If it fits, the usual first step is a two-week Architecture Sprint, which ends with a roadmap, a working proof of concept and a go/no-go recommendation.
Which search and retrieval technologies do you work with?
We have worked with Elasticsearch, Lucene, two-tower retrieval models and RAG pipelines, in prior roles and in consulting work. We choose the approach that fits your data and latency budget, and we assess what you already run before recommending changes.
Do we need to replace our current search system?
Not necessarily. The Architecture Sprint starts by assessing what you have, including a data audit and readiness assessment, and ends with a recommendation on what to change and what to keep.
How do you measure whether relevance improved?
With offline metrics, online experiments and human review, chosen to match what users value in your product. Accuracy alone is rarely the right measure, so defining the metric that tracks user satisfaction is part of the work.
Who owns the work?
You do. You receive the complete codebase, trained models and documentation, with zero lock-in. See How We Work for details.

Tell us what you're building. We'll tell you honestly if we can help.

Book a call
Manmeet Singh
Manmeet SinghFounder · ML Architect