Skip to content
Skip article header Engineering

Shipping a Dense Retrieval Baseline Before Hybrid Search

Our first index used 1024-dimensional dense vectors, with a serving constraint, a versioned collection alias and a deliberately limited retrieval milestone

Updated 5 min read 60 views
A local retrieval workstation showing a passage list beside the selected source page.

Technically reviewed by Victor Sineglazov on .

Our first retrieval diagram included dense vectors, sparse vectors and a reranker. The index we actually built on September 17 used one dense vector per fragment. That difference mattered more than the diagram: a model capable of several retrieval modes does not mean the selected serving path exposes all of them.

I was building the backend for a professional learning platform in aesthetic medicine. Before adding ranking stages, we needed a question to travel through real source fragments and return a citation with a document and page. A smaller search contract let us trace that path without attributing every failure to a hybrid system that did not yet exist.

The serving contract changed the first index

We selected bge-m3 for embeddings and served it through the TEI setup available in our development environment. The initial plan assumed dense and sparse output. Our September implementation notes record that this serving path did not provide the sparse output we needed for that model. We therefore indexed 1024-dimensional dense vectors and used cosine similarity.

This is a historical integration decision, not a claim about every current release of either project. The useful distinction was between a model's advertised capabilities and the response our embedding client could actually consume. A dense-only collection was an honest representation of the latter.

We also left the reranker outside this milestone. Adding it would have introduced another source of changed ordering before we could inspect the basic retrieval result. The first question was whether the expected source reached the candidate set at all.

What the milestone included

The baseline covered the path from a question through vector search, prompt assembly and generation to parsed citations. It also included streamed answers: text could arrive incrementally, with citations and the grounding flag attached to the terminal event. Empty retrieval had a separate behavior: return a refusal without calling the language model.

Stage First implementation boundary What it isolated
PDF ingestion Page-by-page text extraction Whether a fragment retained a source page
Embedding Dense vector, 1024 dimensions The embedding response consumed by the backend
Search Cosine similarity in a versioned collection Candidate retrieval before extra ranking stages
Answer Context fragments and parsed citation markers Whether a returned reference resolved to supplied context
No context Refusal before model invocation Whether an empty search still incurred generation

The first corpus contained 12 public regulatory documents. That was a tractable ingestion scope, not a representative sample of every source the product would eventually handle. Page-oriented text extraction was appropriate for this starting point; it did not establish that diagrams, scanned pages or recordings were covered.

A collection name carried the migration boundary

A backend engineer checking a search collection configuration against a migration checklist.

The application searched through the alias docs. The initial physical collection was docs__bge-m3__v1. Keeping those names separate gave us somewhere to put a later collection with a different vector configuration without making its physical name part of every backend request.

The alias did not make dense and hybrid schemas interchangeable. A future index would still need its own construction and evaluation. It did remove one avoidable coupling: the caller could address a stable logical corpus while the indexing process managed the concrete collection behind it.

Indexing lived inside the Java backend. The same ownership boundary covered the embedding client and vector-store writes, rather than splitting corpus creation into a second application with a separate interpretation of fragment metadata. The indexing command read a corpus manifest, created fragments and switched the alias after building the collection.

What we deliberately deferred

The September milestone excluded chat history, authentication and frontend work. PostgreSQL was present in the development environment but was not yet part of this backend path. A richer document parser was also deferred. Those exclusions gave the milestone a concrete finish line: source text in, question in, cited answer or refusal out.

Deferral still had costs. Dense search alone gave us no direct evidence about the benefit of sparse matching for specialized identifiers. Omitting a reranker meant the first ordering came from vector similarity. We could examine those limitations because they were explicit, rather than silently describing the baseline as the full architecture.

The evaluation target in the design was recall@5 of at least 0.8 without a device filter. A target is not a result. It defined one acceptance question for the retrieval stage and could not certify answer correctness, clinical suitability or performance on later source types. Our separate article on a cutoff that failed after the corpus changed follows that boundary further.

The evidence we needed before adding another stage

For each missed answer, the baseline offered a short diagnostic path: inspect extracted text, inspect candidate retrieval, inspect the supplied prompt context and then inspect the returned citation. A sparse branch or reranker would add useful behavior only if we could show which failure it addressed. Otherwise, a better-looking answer could conceal an ingestion defect or a reference that merely looked valid.

That distinction is especially relevant to our work on a swappable model interface. Changing the generator cannot recover a source fragment that never reached its context. The first collection gave us a place to observe that absence.

Evidence scope: this account uses the September 17 retrieval design and the September 17 implementation correction in our R&D pipeline record. It describes the first dense index and its stated acceptance boundaries. It makes no new comparison of dense and hybrid retrieval quality.

FAQ

Last updated:

Quick answers to common questions about custom software development, pricing, process and technology.

  • How long have you been in business?

    Pharos Production has been in business since 2013, with over 13 years of experience in custom software development. During this time, we have delivered over 110 applications for 200+ clients across 28 industries, including FinTech, healthcare, crypto and e-commerce. We are rated 5/5 on Clutch based on 107 verified reviews (2026).

  • What services do you provide?

    Pharos Production provides six core service categories: AI Development (LLM applications, AI agents, machine learning and computer vision), Blockchain and Web3 (smart contracts, DeFi, tokenization on Ethereum, Solana, TON and other chains), FinTech Development (payment platforms, digital banking, KYC and AML integration), Software Development (mobile apps, web platforms, database design, UI/UX), Security and QA (code audits, penetration testing, smart contract audits, manual and automated testing) and Consulting (architecture design, MVP validation, startup consulting).

  • Where is Pharos Production located?

    Pharos Production is headquartered in Las Vegas, Nevada, USA (5348 Vegas Dr, Las Vegas, NV 89108), with an engineering office in Kyiv, Ukraine (44-B Eugene Konovalets Str. Suite 201, Kyiv 01133). We work with clients worldwide and provide remote collaboration across all time zones. Visit our contact page for directions and scheduling options.

  • How big is your team?

    Pharos Production has a team of 90+ engineers, including software developers, blockchain specialists, QA engineers, DevOps experts, UI/UX designers, project managers and solution architects. Our founder, Dr. Dmytro Nasyrov, holds a PhD in Artificial Intelligence and leads the technical direction of all projects.

  • What types of customers do you serve?

    We serve a wide range of clients, from startups and product companies to mid-sized enterprises and large institutions. Our clients include crypto exchanges, FinTech providers (like Pleenk), healthcare organizations, sportsbook operators (like Pro Gambling), e-commerce platforms and SaaS companies. Pharos Production has worked with 200+ clients across 28 industries since 2013, adapting engagement models to match each client's stage, whether it is MVP validation for a startup or enterprise-scale development for an established business.

  • How is a custom software development company different from an agency?

    A custom software development company like Pharos Production assigns one dedicated engineering team that owns your product across multiple release cycles, instead of handing it off once a contract ends. Agencies more often rotate staff between clients and optimize for fast delivery over long-term product ownership.

    Pharos Production keeps the same team on your codebase through 2-week sprints, so there is no re-onboarding cost each time you request a new feature.

  • How much does custom software development cost?

    Custom software development costs vary based on project scope and complexity. At Pharos Production, typical project ranges are: MVP development ($10,000-$25,000), suitable for startups validating a product idea; full-fledged production ($25,000-$50,000), for established businesses scaling a proven concept; and full-cycle development ($50,000-$80,000+), for complex enterprise-grade systems. These ranges include architecture design, development, QA testing and deployment. Hourly rates range from $50 to $99 depending on role and seniority. Final pricing depends on technology stack, number of integrations and engagement model (staff augmentation, dedicated team or project outsourcing).

  • How long does it take to develop a custom software application?

    Development timelines depend on scope and complexity. At Pharos Production, a typical MVP takes 2-4 months, a production-ready application takes 4-8 months and a complex enterprise system can take 8-12+ months. We use an agile methodology with 2-week sprints, delivering working increments after each sprint. Every sprint includes a retrospective, progress report and planning session for the next iteration. This approach ensures transparency and allows businesses to launch faster by prioritizing high-impact features first. Get a timeline estimate for your project.

  • What industries does Pharos Production serve?

    Pharos Production serves 28 industries: FinTech, MiCA Compliance, Banking, Insurance, Crypto, Healthcare, Fitness, E-Commerce, Food & Restaurant, On-Demand, Manufacturing, Automotive, Energy, Oil & Gas, Supply Chain, Construction, Real Estate, Gaming, Streaming, Media, Telecom, Education, Aviation, Travel, Logistics, Marketing, Agriculture and Recommender Systems. Our deepest expertise is in FinTech, blockchain and healthcare, where we have delivered compliance-ready platforms (HIPAA, PCI DSS, GDPR) and high-load systems handling thousands of concurrent users. For the latest industry insights, read our guides on FinTech trends in 2026 and the Web3 technology stack.

  • Why hire a software development company instead of building in-house?

    Hiring a software development company offers faster time-to-market, lower upfront costs and access to specialized expertise without long-term employment commitments. According to Deloitte's 2024 Global Outsourcing Survey, 57% of companies outsource software development to access skills they cannot hire internally.

    Factor In-house team Software development company
    Time to assemble 3-6 months (recruiting + onboarding) 1-2 weeks
    Upfront cost High (salaries, benefits, equipment) Lower (project-based pricing)
    Specialized expertise Limited to who you can hire locally Access to 90+ engineers across blockchain, AI, FinTech
    Scalability Slow (each new hire takes months) Fast (scale up or down per sprint)
    Long-term commitment Full-time employment contracts Flexible engagement models
    Risk High if key engineers leave Company ensures continuity and knowledge transfer

    For businesses that need blockchain, AI or high-load architecture expertise, outsourcing to a specialized firm like Pharos Production reduces risk and accelerates delivery.

  • What projects does Pharos Production not take on?

    Pharos Production focuses on mid-to-large custom software projects with budgets starting at $10,000. We do not take on template-based websites, WordPress theme customization or short-term contracts under one month. We also do not provide non-technical staffing (marketing, sales or design-only roles). Our strongest fit is blockchain, FinTech and healthcare projects where security, compliance and high-load architecture are critical. For smaller projects or MVPs under $10,000, we recommend exploring freelance platforms or no-code tools as a more cost-effective starting point.

  • Why does Pharos Production use agile methodology?

    We use agile with 2-week sprints because it reduces the risk of building features that miss the mark. Each sprint ends with a working demo, a retrospective and a plan for the next iteration.

    This means clients see progress every 14 days and can adjust priorities based on real results, not assumptions. We chose this approach after years of experience showing that rigid, fixed-scope contracts lead to scope creep, missed deadlines and products that do not match market needs by launch day.

  • When should you NOT hire a custom software development company?

    Custom development is not the right choice in every situation. You should not hire a custom software company if: your problem is fully solved by an existing SaaS product (e.g. Shopify for e-commerce, Salesforce for CRM); your budget is under $10,000 and timeline is under 4 weeks; you need a simple landing page or marketing website (WordPress or Webflow is faster and cheaper); or you are still validating the idea and have not spoken to potential users yet.

    In these cases, off-the-shelf tools or no-code platforms offer better ROI. Custom development makes sense when you need unique workflows, regulatory compliance, high-load architecture or competitive differentiation that packaged software cannot provide.

  • What kind of projects has Pharos Production delivered?

    Here are three anonymized examples from our recent delivery history:

    FinTech startup - payment platform (MVP)
    Scope: mobile app + backend API with bank-grade encryption. Team: 4 engineers, 1 QA. Timeline: 10 weeks. Budget: $38,000. Result: launched on schedule, processed $2M+ in transactions within the first quarter.

    Healthcare provider - patient portal (Full product)
    Scope: HIPAA-aligned web platform with EHR integration, appointment scheduling and telemedicine. Team: 6 engineers, 1 DevOps, 2 QA. Timeline: 6 months. Budget: $120,000. Result: 15,000+ active patients, zero compliance violations in two annual audits.

    Crypto exchange - trading engine (Complex)
    Scope: high-load matching engine handling 50,000+ orders per second, multi-chain wallet infrastructure on Ethereum and Solana. Team: 8 engineers, 2 QA, 1 security auditor. Timeline: 11 months. Budget: $340,000. Result: 99.97% uptime, passed three independent security audits.

    See more projects: NoMoreBets, Pulse, Sagas, Gambit Stream and Pleenk. For the full portfolio, visit our case studies. Learn more about the technology behind these projects in our guide to stablecoins and crypto infrastructure.

Dmytro Ihnatov is the Lead Backend Developer at Pharos Production. He documents backend architecture and research experiments from the team's work on a professional learning platform for aesthetic medicine.

Dmytro Nasyrov, Founder and CTO at Pharos Production
Dmytro Nasyrov Founder & CTO Let's work together!

Your business results matter

Achieve them with minimized risk through our bespoke innovation capabilities

Your contact details
Please enter your name
Please enter a valid email address
Please enter your message

We use your details only to reply to your request. Data Privacy and Legal Notice

We typically reply within 24 hours

What happens next?

  1. Contact us

    Contact us today to discuss your project. We're ready to review your request promptly and guide you on the best next steps for collaboration

    Same day
  2. NDA

    We're committed to keeping your information confidential, so we'll sign a Non-Disclosure Agreement

    1 day
  3. Plan the Goals

    After we chat about your goals and needs, we'll craft a comprehensive proposal detailing the project scope, team, timeline and budget

    3-5 days
  4. Finalize the Details

    Let's connect on Google Meet to go through the proposal and confirm all the details together!

    1-2 days
  5. Sign the Contract

    As soon as the contract is signed, our dedicated team will jump into action on your project!

    Same day