Design

What is design??

RAG pipeline – Cost?

“Vector DB + LLM = Done!”

Incorporate some open-source tools and include Langchain (we’ll discuss that later). You should be all set, right?

Wrong.

Here’s the problem:

  • One full-time engineer is dedicated to debugging hallucinations and accuracy issues.
  • One full-time data specialist is responsible for dealing with ETL and data ingestion problems.
  • One full-time DevOps engineer needs help with scalability and infrastructure challenges.
  • One very frustrated CTO is facing a tripled budget.

Here are some factors that were not considered:

  1. Complexity of Document and Knowledge Base Pre-Processing: This includes the challenges of ingesting various data sources such as SharePoint, Google Drive, and websites.
  2. Document Formats and PDF Issues: There are various complications that arise when importing different formats, including PDFs and EPUB files.
  3. Accuracy Issues in Production: While everything may work well in testing, actual production usage may present problems, especially when used by real users.
  4. Hallucinations: This refers to instances when the system generates incorrect or nonsensical information.
  5. Response Quality Assurance: Ensuring the quality and reliability of responses generated by the system.
  6. Integration with Existing Systems: Compatibility and seamless integration with current systems should be considered.
  7. Change Data Capture: Maintaining synchronization with data changes on websites is essential.
  8. Compliance and Audit Requirements: Adhering to relevant legal and regulatory standards.
  9. Security Issues and Data Leaks: Ensuring that your internal system meets SOC-2 Type 2 compliance and is secure from potential data breaches.

The Cost Nobody Talks About

Here’s a breakdown of the true costs associated with your “free” RAG (Retrieval-Augmented Generation) system:

Infrastructure Costs

  • Hosting for vector databases
  • Costs of model inference
  • Development environments
  • Testing environments
  • Production environments
  • Backup systems
  • Monitoring systems

Personnel Costs

  • ML Engineers: $150,000 – $250,000 per year
  • DevOps Engineers: $120,000 – $180,000 per year
  • AI Security Specialists: $160,000 – $220,000 per year
  • Quality Assurance: $90,000 – $130,000 per year
  • Project Manager: $100,000 – $200,000 per year

Ongoing Operational Costs

  • 24/7 monitoring
  • Security updates
  • Model upgrades
  • Data cleaning
  • Performance optimization
  • Documentation updates
  • Training for new team members
  • Compliance audits
  • Maintaining feature parity as AI evolves

This breakdown highlights the various costs involved in running a RAG system, illustrating that it may be far from free.

The Security Nightmare

Want to lose sleep? Try being responsible for an AI system that:

  • Has access to your company’s entire knowledge base
  • Could potentially leak sensitive information
  • Might generate inaccurate or misleading information about confidential data
  • Requires constant security updates
  • Could be vulnerable to prompt injection attacks
  • Might unintentionally reveal internal data through its responses
  • Could be susceptible to adversarial attacks

Managing such an AI system can be a daunting challenge.

Think about this: every new document you add to your knowledge base can pose a potential security risk. Every prompt you receive is a possible attack vector, and every response must be carefully screened. It’s not only about creating a secure system—it’s also about maintaining that security in an environment that is constantly changing.

Daily Maintenance Tasks:

  • Monitor response quality
  • Check for hallucinations
  • Debug edge cases
  • Handle data processing issues
  • Manage API quotas and infrastructure concerns

Weekly Maintenance Tasks:

  • Optimize performance
  • Conduct security audits
  • Perform data quality checks
  • Analyze user feedback
  • Implement system updates

Monthly Maintenance Tasks:

  • Conduct large-scale testing
  • Update AI models
  • Review compliance
  • Optimize costs
  • Plan for capacity
  • Review system architecture
  • Align with strategic goals
  • Address feature requests

The Expertise Gap

ML Operations

  • LLM Model deployment expertise
  • RAG pipeline management
  • Version control for models
  • Accuracy optimization
  • Resource management
  • Scaling knowledge

RAG Expertise

  • Understanding accuracy
  • Anti-hallucination optimization
  • Context window optimization.
  • Understanding latency and costs.
  • Prompt engineering
  • Quality metrics

Infrastructure Knowledge

  • Vector database optimization
  • Logging and monitoring.
  • API management
  • Cost optimization
  • Scaling architecture

Security Expertise

  • AI-specific security measures
  • Prompt injection prevention
  • Data privacy management
  • Access control
  • Audit logging
  • Compliance management

The Time-to-Market Reality

While you’re developing your RAG system:

  • Your competitors are launching production solutions.
  • Technology is evolving, sometimes on a weekly basis.
  • Your requirements are continually changing.
  • Your business risks losing opportunities.
  • The market is advancing rapidly.
  • Your initial design is becoming outdated.
  • User expectations, influenced by OpenAI, are rising daily.

Month 1: Initial Development

  • Establishing basic architecture
  • Creating the first prototype
  • Conducting initial testing
  • Gathering early feedback

Month 2: Facing Reality

  • Identifying security issues
  • Uncovering performance problems
  • Encountering an increase in edge cases
  • Adapting to changing requirements

Month 3: Rebuilding

  • Revising the architecture
  • Enhancing security measures
  • Optimizing performance
  • Catching up on documentation

Month 4: Preparing for Enterprise Readiness

  • Implementing compliance measures
  • Setting up monitoring systems
  • Planning for disaster recovery
  • Providing user training

The Buy Alternative

Modern RAG Solutions Offer:

Infrastructure Management

  • Scalable architecture
  • Automatic updates
  • Performance optimization
  • Security maintenance

Enterprise Features

  • Role-based access control
  • Audit logging
  • Compliance management
  • Data privacy controls

Operational Benefits

  • Expert support
  • Regular updates
  • Security patches
  • Performance monitoring

Business Advantages

  • Faster time-to-market
  • Lower total cost
  • Reduced risk
  • Proven solutions

When Should You Build?

There are three specific scenarios where building your own solution makes sense:

  1. Unique Regulatory Requirements:
  • Custom government regulations
  • Specific industry compliance needs
  • Unique security protocols
  1. Core Product Development:
  • It serves as your main value proposition
  • You are innovating in the space
  • You possess deep expertise
  1. Unlimited Resources:
  • If you truly have unlimited time and money (though this situation is rare)

Even with ample resources, opportunity cost matters, and time-to-market is important.


Here’s What You Should Do Instead:

  • Focus on your actual business problems:
  • What are your users trying to achieve?
  • What are your unique value propositions?
  • Where can you make the biggest impact?
  • Choose a Reliable RAG Provider:
  • Evaluate based on your specific needs (Hint: Review case studies)
  • Check security certifications (Hint: Look for SOC-2 Type 2)
  • Verify enterprise readiness (Hint: Request case studies!)
  • Test performance (Hint: Look for published benchmarks)
  • Assess support quality (Hint: Contact support!)
  • Dedicate your engineering efforts to areas that truly differentiate your business:
  • Custom integrations
  • Unique features
  • Business logic
  • User experience

The Bottom Line

Stop trying to reinvent the wheel.

High-Definition Colors

We are currently experiencing exciting developments in the realm of color. With advanced high-definition color spaces like LCH, OKLCH, LAB, and OKLAB, we now have access to 50% more colors available in modern browsers, which may soon render the traditional RGB and HSL formats obsolete.

To help you become familiar with these new color spaces, it’s important to note that both OKLCH and OKLAB are based on human perception and can accurately represent any color visible to the human eye. OKLAB is particularly effective for creating rich gradients, while OKLCH is an excellent choice for developing color palettes in design systems. Notably, OKCLH and OKLAB colors are fully supported in Chrome, Edge, Safari, Firefox, and Opera.

  1. sRGB is no longer the best color space
    The sRGB color space covers only 30% of the colors that the human eye can perceive. Color definitions in RGB and HSL are limited to sRGB. However, we now have the option to use the display-p3 color space, which encompasses 50% more colors.
  2. OKLCH and OKLAB are based on human perception
    In HSL, lightness and saturation each range from 0% to 100%, but this does not align well with how our eyes perceive light. OKLCH considers these specifics, making colors appear more authentic to our perception.
  3. Understanding the OKLCH Syntax
    OKLCH stands for Lightness, Chroma, and Hue (LCH). The syntax for OKLCH is structured as oklch(100% 0.37 330 deg). Lightness ranges from 0% to 100%, while Chroma is a value between 0 and 0.37, indicating the color’s intensity. Most hues would set the upper limit for Chroma at 0.187.
  4. Differences in Hue Values
    The hue values in OKLCH differ from those in HSL, as all hues have been shifted by approximately 30 degrees due to perceptual differences. This means that lightness cannot be translated 1:1 from HSL to OKLCH.
  5. Optimizing Gradients with OKLAB
    OKLAB is tailored for superior image processing quality and serves as the default color space for animations and interpolations. It is particularly effective for creating perceptually consistent gradients and has a high dynamic range.
  6. Using OKLCH for Design Systems
    OKLCH is ideal for developing color palettes in design systems. Similar to HSL, you can create shades of colors by adjusting the color intensity. Unlike RGB, there is no need for guesswork, and the syntax is more readable and user-friendly.
  7. Comprehensive Color Specification
    With OKLCH and OKLAB in the display-p3 color space, we can define any color our eyes can perceive. Colors that fall outside of the supported range are rounded to the nearest available color.
  8. Browser Support for OKLCH/OKLAB
    OKLCH and OKLAB are fully supported in modern browsers, including Chrome, Edge, Safari, Firefox, and Opera. Feature queries in CSS can be used to target browsers that support these color spaces, while providing fallbacks for those that do not.

Currently, OKLCH and OKLAB colors are not available in Figma, but when they are, we will experience a transformative approach to using color on the web.

About input mode

The inputmode global attribute provides a hint to browsers for devices with onscreen keyboards to help them decide which keyboard to display when a user has selected any input or textarea element.

<input type="text" inputmode="" /> <textarea inputmode="" />

Nearly six years later, Chrome for Android added the feature, and with the recent release of iOS 12.2, Safari supports it as well.

inputmode accepts a number of values. Let’s go through them, one by one.

None

<input type="text" inputmode="none" />

Numeric

<input type="text" inputmode="numeric" />

Tel

<input type="text" inputmode="tel" />

Decimal

<input type="text" inputmode="decimal" />

Email

<input type="text" inputmode="email" />

URL

<input type="text" inputmode="url" />

Search

<input type="text" inputmode="search" />

Issues with HR Recruiting Software

This article may become outdated in a few months as they might be fixing the updates for those issues. An Applicant Tracking System (ATS) is software that reads and sorts resumes for HR purposes. However, ATS has limitations:

  • It cannot read images, columns, tables, or semicolons.
  • It does not recognize PDF format.

To ensure your resume is ATS-friendly, consider the following tips:

  1. Remove any images.
  2. Eliminate tables and semicolons.
  3. Use only text formatting with letters, commas, periods, and bullet points.
  4. Stick to black font color.
  5. Limit your document to one page.
  6. Rename the document using only your last name.
  7. Change the format from PDF to DOC.
  8. Include a skills section featuring specifically selected keywords.
  9. Use Times New Roman font, size 12.

Following these guidelines can improve your chances of passing through the ATS effectively.


[Your Name]
[Your City, State] • [Your Phone Number] • [Your Email Address] • [LinkedIn URL or Portfolio URL]

Professional Summary
A concise summary (2-3 sentences) that highlights your experience, skills, and specific roles or industries. Tailor this section to the job description, incorporating relevant keywords directly from the posting.

Skills

  • Core Skills: List your top skills relevant to the job, such as “Project Management,” “Data Analysis,” or “Digital Marketing.”
  • Technical Skills: Mention specific software, platforms, or tools (e.g., Microsoft Office Suite, Adobe Creative Cloud, SQL, Python).
  • Soft Skills: Include key interpersonal skills (e.g., leadership, communication) that are relevant to the position.

Professional Experience

Job Title
Company Name, Location (City, State)
Dates of Employment (Month Year – Month Year)

  • Responsibilities: Use bullet points to outline key responsibilities, focusing on quantifiable achievements (e.g., “Increased sales by 20% by developing targeted marketing campaigns”).
  • Achievements: Include metrics whenever possible (e.g., “Reduced project turnaround time by 15%”).

(Repeat for each relevant job in reverse chronological order.)

Education
Degree (e.g., Bachelor of Science in Marketing)
University Name, Location (City, State)
Graduation Year (or “Expected [Year]” if still studying)

Certifications (if applicable)
Certification Title, Issuing Organization, Year Awarded (e.g., “Project Management Professional (PMP), PMI, 2022”)

Additional Sections (optional)

  • Volunteer Experience: If relevant to the position.
  • Languages: List languages and proficiency (e.g., “Spanish (Fluent)”).
  • Professional Affiliations: Any industry associations.

Tips for ATS Optimization

  • File Format: Save as a .docx or .pdf (if the system allows).
  • Consistent Formatting: Use clear section headings, standard fonts (Arial, Times New Roman, Calibri), and avoid images, tables, or special characters.
  • Relevant Keywords: Ensure that skills and job-specific terms align with the job description.

a Simple CSS One-liner

The Simple CSS Code for Background Blur

backdrop-filter: blur(6px);

See the Pen Blurred by El Diablo (@diablo) on CodePen.

Page 1 of 23

Powered by WordPress & Theme by Anders Norén