Letters of Recommendation: How to Write, Ask for, and Format One
By
Liz Fujiwara
•
Mar 4, 2026
Picture this: it is mid-2026, and you are an AI engineer applying for a Staff ML role at a Series B startup and an AI residency at a major research lab. Your GitHub is polished and your arXiv preprints are gaining citations, but both applications still require something no portfolio can replace, a strong, human-backed letter of recommendation.
A letter of recommendation is a signed, dated document authored by someone who knows your professional or academic performance and can speak to your skills, experience, and character, providing validation beyond what a resume or coding assessment can show. While your GitHub, Kaggle medals, and publications demonstrate what you have built, a recommender provides insight into how you work, including reliability under pressure, collaboration style, mentorship, and actual impact on projects and teams.
Key Takeaways
A letter of recommendation is a formal document where a credible professional or academic contact endorses your skills, character, and track record, and it remains essential for AI engineers, ML researchers, and infrastructure specialists competing for top roles in 2026.
This guide covers how to ask for letters, how to write them as a recommender, how to format them for maximum impact, who to ask, ideal timelines, and what makes a letter stand out to hiring committees at FAANG+ companies and AI labs.
Platforms like Fonzi use AI to reduce noise and bias without replacing human judgment, turning your recommendations and project history into high-signal inputs, and Match Day provides a curated, time-bound opportunity to connect with serious employers efficiently.
What Is a Letter of Recommendation in Today’s AI Job Market?
A letter of recommendation is a formal document where a credible professional or academic contact endorses your skills, character, and track record for a specific opportunity. It is signed, dated, and typically submitted directly to the requesting organization, whether that is a hiring committee, admissions office, or fellowship program.
The purpose and emphasis of letters differ based on context:
For industry roles (e.g., Senior ML Engineer at a Series B AI startup in 2026), letters focus on your ability to ship production systems, collaborate across teams, and deliver measurable results.
For research roles (e.g., AI residency at a major lab), letters emphasize intellectual curiosity, research rigor, and your contributions to published work or ongoing projects.
For grad school or PhD applications, letters provide admissions officers with insight into your academic potential, work ethic, and fit for the particular program.
In AI-heavy hiring, letters often confirm things that technical screens cannot, such as code quality over time, ability to ship models to production under real constraints, research rigor beyond a single paper, and collaboration across product, research, and infrastructure teams.
Recommenders are usually prior managers, tech leads, founders, PIs, or professors who have worked with you closely for at least one substantial project or a three to six month period. The depth of the relationship matters more than the seniority of the recommender.
A strong letter complements modern screening, including ATS filters, coding assessments, take-home tasks, and structured interviews, rather than replacing them. It provides human-validated context that algorithms and test scores alone cannot capture.
Who Should You Ask for a Letter of Recommendation?
The strength of a letter depends more on depth of relationship and observed performance than on job title alone. A staff engineer who supervised your work daily will write a more compelling letter than a VP who barely knows your name.
Prioritize these recommenders for AI/ML roles:
Managers at your last AI startup who can speak to your day-to-day contributions and growth
Staff or senior engineers who supervised your infrastructure or model development work
Research advisors on your papers who can address your intellectual contributions
Open-source project maintainers who have reviewed your PRs and observed your collaboration style
When to choose specific types of recommenders:
An academic recommender (e.g., thesis advisor on a 2024 RL or LLM thesis) is ideal for PhD or research roles where intellectual rigor is paramount.
An industry recommender (e.g., Head of ML Platform) is best for senior engineering roles at product companies where shipping matters.
A hybrid profile (someone who moved from research to industry) works well for AI residency and applied research teams that value both perspectives.
Avoid these recommenders:
Very senior people who barely know your work and will write a generic letter
Personal friends who cannot provide specific examples of professional performance
Anyone who seems hesitant, noncommittal, or unable to enthusiastically support your application
Keep a simple “recommenders log” listing people you have worked with, specific projects and dates, and which of your strengths each person can speak to. This makes future requests faster and more strategic.
What to Prepare Before Requesting a Recommendation Letter
Good preparation makes it easier for recommenders to write detailed, specific letters quickly and significantly increases the quality of what they produce.
Key materials to send your recommender:
A current resume or CV focused on AI and infra projects, with dates and measurable results
A 1-page brag sheet summarizing 3-5 concrete achievements (e.g., “Reduced inference latency by 40% on production LLM serving stack,” “Co-authored paper accepted to NeurIPS 2024”)
Links to GitHub, major repos, arXiv papers, Kaggle profiles, and significant blog posts or talks
Your transcript if applying to academic programs
Include a short paragraph covering:
What you’re applying for (e.g., “Applied Scientist, Amazon, start date Q4 2026”)
Why you’re a strong fit (2-3 bullet points tailored to the job description)
Any specific topics the letter should address (teamwork, mentoring, research independence, infra reliability)
Provide all the details your recommender needs, including submission instructions such as portal links, email addresses, or PDF requirements, and the application deadline. Schedule a 20 to 30 minute call or in-person meeting so the recommender can ask clarifying questions and align on deadlines and target roles. This conversation often surfaces details and stories that improve the letter.
When to Ask: Timelines and Deadlines for 2026 Applications
Early, respectful requests lead to better letters. Recommenders, especially busy professors and tech leads, often cap the number of letters they will write per season and prioritize candidates who plan in advance.
Concrete timing guidance:
For fall 2026 graduate programs with December deadlines, ask by August or early September 2026; at least 6-8 weeks before letters are due.
For AI residency programs with annual deadlines (typically October-December), request letters 6 weeks before the deadline.
For industry roles where reference checks happen in final stages (e.g., senior AI engineer hiring in Q3 2026), ask 3-4 weeks before you expect referral checks or committee review.
Many professors and tech leads receive numerous requests during peak application season. Asking early rather than making a last-minute request shows professionalism and respect for their time. If you are switching jobs quickly due to layoffs or acqui-hires in 2026, request at least one “evergreen” letter addressed “To whom it may concern” that can be adapted later for specific opportunities.
How to Ask for a Letter of Recommendation (Without Making It Awkward)
Even experienced engineers often feel awkward asking for letters. The good news is this is a completely standard part of professional life, and people who have worked with you typically expect these requests.
Best practice is to ask synchronously first, via video call or in person, and then follow up with an email summarizing details, deadlines, and links.
What to say in a live request:
Remind them how you worked together: “We collaborated on the 2023 recommender system revamp, and I really valued your mentorship during that project.”
Share what you’re applying to and why you’re excited about the opportunity.
Ask directly if they’d feel comfortable writing a strong, positive recommendation: “Would you feel comfortable writing a strong letter supporting my application?”
Be explicit about timing: “The deadline is 15 October 2026, and they’d need the letter one week before.”
If someone says no or seems lukewarm, accept it gracefully. A hesitant recommender will write a lukewarm letter, which can hurt your application more than help it. Thank them and move on to someone else. Give recommenders an easy way to decline, for example, “I completely understand if your schedule is full or if you do not feel you know my work well enough.” This removes awkwardness and helps you get honest answers.
Recommendation Request Email Example (For Technical Roles)
After your live conversation, send a follow up email that summarizes everything the recommender needs. This email should be concise (200-300 words), professional, and easy to reference when they sit down to write.
Ideal structure for the email:
Subject line: Clear and time-bound (e.g., “Letter of Recommendation Request for PhD Applications – Due 1 Dec 2026”)
Opening line thanking them for agreeing to write and briefly recalling how you’ve worked together
2-3 sentence description of the roles or programs (e.g., “Research Engineer at OpenAI, start date summer 2026; focus on safety tooling”)
Bullet list of attachments and links: resume, brag sheet, portfolio, GitHub, publications
Explicit deadline and submission instructions (upload portal link, email address, or PDF requirements)
A clear sentence giving them an easy out: “I completely understand if your schedule changes and you’re no longer able to help.”
Closing line expressing genuine appreciation with your contact details
Keep the tone neutral and professional. Avoid over-explaining or adding unnecessary context, as your recommender has already agreed to help. You can use this structure as a template for multiple requests, customizing a few sentences for each recommender based on your specific relationship and the projects you worked on together.
How to Write a Letter of Recommendation (If You’re the Recommender)
Many readers will eventually be asked to write letters for peers, reports, or mentees. This section helps you do so responsibly and effectively.
Standard structure for a strong recommendation letter:
Introduction (1-2 sentences): State your role and relationship to the candidate. Example: “I am the Director of ML at X, and I managed Y from January 2023 to June 2024.”
Overview (1 paragraph): Describe the candidate’s role and scope. Example: “Y led the redesign of our LLM inference stack handling 10M daily requests.”
Strengths (2-3 paragraphs): Each paragraph should center on a specific strength with concrete evidence; technical depth, reliability, communication, leadership/mentoring.
Peer comparison (1 paragraph): Compare the candidate to peers. Example: “Y is in the top 5% among engineers I’ve managed in the last 10 years.”
Closing (1 paragraph): Provide an unambiguous recommendation and your contact info. Example: “I recommend Y without reservation and would be happy to hire them again.”
Include specific examples with metrics such as latency reductions, cost savings, uptime improvements, or research outputs with publication venues and years. Scholarship committees and employers respond to evidence, not vague praise.
Honesty matters. Do not exaggerate or fabricate achievements. If you cannot be strongly positive about the candidate, it is better to decline the request than to write a lukewarm letter that will hurt their chances.
Aim for at least one page for most industry roles and one to two pages for research or academic recommendations, unless the program specifies otherwise.
Formatting a Letter of Recommendation: Structure, Style, and Details
Consistent formatting helps letters be taken seriously by hiring committees and admissions officers. A well-formatted letter signals professionalism and attention to detail.
Required elements:
Date at the top (e.g., “24 February 2026”)
Recipient name, title, organization, and address if known; otherwise “To whom it may concern”
A clear subject line for email-based submissions (e.g., “Recommendation for [Full Name] – Senior ML Engineer”)
Business or organization letterhead if available (or at minimum, your professional contact information)
Layout guidelines:
Use professional fonts (11-12 pt serif or sans-serif)
Single-spaced paragraphs with blank lines between paragraphs
Formal salutation (e.g., “Dear Hiring Committee,”)
Formal closing (e.g., “Sincerely,”)
Typed name and title, with optional scanned signature
Digital submission details:
Save as PDF unless otherwise specified
Include the candidate’s name in the file name (e.g., “LOR_Alex_Kim_Google_Research_2026.pdf”)
Follow specific portal instructions regarding character limits, organization letterhead requirements, or text boxes
Submit the recommendation directly through the portal rather than sending to the candidate
Anatomy of a Strong Recommendation Letter vs. a Weak One
Understanding what distinguishes a good letter from a generic one helps both candidates and recommenders. Use this table to quickly see what committees are actually looking for in 2026.
Dimension | Strong Letter Element | Weak Letter Element |
Specificity of Impact | “Reduced inference latency by 40% on our production LLM serving stack in Q2 2024” | “Is a hard worker who contributes to the team” |
Knowledge of Candidate | “I worked directly with Alex for 18 months as their tech lead on three major projects” | “I am familiar with their work through occasional meetings” |
Comparison to Peers | “Top 5% among the 40+ ML engineers I’ve managed over my career” | “A strong candidate for the role” |
Alignment with Target Role | “Alex’s experience scaling transformer inference makes them ideal for this LLM infrastructure position” | Generic praise that could apply to any role |
Clarity of Endorsement | “I recommend Alex without reservation and would hire them again immediately” | “I think Alex would be a decent fit” |
Concrete Evidence | References specific projects, metrics, and student’s strengths with examples | Uses vague adjectives without evidence |
Understanding of Context | Tailored to the specific company, program, or role being applied to | Reads like a form letter sent to multiple recipients |
Use this as a recommendation template when drafting or reviewing letters. The strongest letters address multiple rows in the “Strong” column with specific examples drawn from direct observation.
Using AI Tools to Draft or Support Recommendation Letters (Responsibly)
Large language models can help structure or polish letters, but they must not replace human judgment or firsthand experience. Ethical AI use in recommendations requires clear boundaries.
Safe use cases for AI tools:
Turning bullet notes about a candidate’s achievements into a first-draft paragraph
Checking grammar, clarity, and tone
Adjusting length and formality for different recipients (startups vs. large enterprises vs. grad committees)
Suggesting stronger action verbs or helping explain technical work for non-technical readers
What AI tools must never do:
Invent projects, metrics, or credentials that don’t exist
Impersonate the recommender without their review and final edits
Bypass institutional policies on AI-assisted writing
Generate entire letters without human oversight and verification
Fonzi’s use of AI is transparent. Algorithms help surface patterns in candidate profiles and job needs, but every match is reviewed by humans who understand the technical stack. This approach protects both candidates and employers. AI in hiring, when done responsibly, should reduce repetitive tasks and noise so humans can focus on work and potential. The goal is efficiency with integrity, not automation without accountability.
How Companies Use AI in Hiring and How Fonzi Is Different
Most companies now use AI somewhere in their hiring process, including resume parsing, keyword matching, coding test autograding, and candidate prioritization based on historical patterns. Recent reports indicate about 60 percent of tech firms blend traditional evaluation methods, like letters, with AI screening.
This creates specific risks:
Over-reliance on keyword scoring that penalizes non-traditional backgrounds or research-heavy resumes
Biased historical data that perpetuates existing hiring patterns
Opaque filters that reject strong candidates before humans ever see their applications
Fonzi’s marketplace works differently:
Curated candidate pool: Invite-only or screened candidates focused on AI engineers, ML researchers, infra engineers, and LLM specialists
Human expert review: Profiles, projects, and signals (including recommendations) are reviewed before matching
AI as augmentation: AI tools highlight fit between candidates and roles (experience with specific frameworks, infra stacks, or research subfields) rather than auto-rejecting
High-signal inputs: Letters of recommendation and project endorsements are incorporated as meaningful inputs into matching, not binary pass/fail gates
Fonzi reduces bias through standardized profiles, consistent evaluation rubrics, and emphasis on verified skills and contributions over superficial markers. The result is a faster and fairer process for everyone involved.
Inside Fonzi’s Match Day: Turning Recommendations into Real Opportunities
Match Day is a specific, time-bound event where pre-vetted AI and infra candidates are presented simultaneously to multiple vetted companies. It’s designed to create high-signal connections efficiently.
The process works like this:
Candidates apply once and complete a structured profile including projects, publications, and recommendation signals
Fonzi reviews and curates a set of candidates for a given Match Day (e.g., one focused on LLM infrastructure in Q3 2026)
Companies receive structured candidate briefs on Match Day and respond with interview requests within a tight window
Benefits for candidates:
No endless cold outreach; high-signal exposure to hiring teams already budgeted and aligned to hire
Faster feedback loops; often hearing back in days rather than weeks
Greater transparency on role expectations, tech stacks, and process
Well-prepared letters of recommendation make a tangible difference in how candidates are prioritized within Match Day cohorts. When companies see strong endorsements alongside project evidence, they are more likely to prioritize those candidates for interview slots.
This is what modern hiring looks like when AI and human judgment work together: efficiency without sacrificing the personal elements that predict success.
Practical Tips for Showcasing Your Skills Alongside Recommendations
Letters of recommendation are one part of a broader evidence set that includes code, research, systems you’ve shipped, and your communication skills. The strongest applications tell a coherent story across all these elements.
Tips for AI engineers and ML researchers:
Keep a living portfolio (GitHub, personal site) with clearly documented projects and impact summaries
Write short case studies for 2-3 major projects, including architecture diagrams, metrics, and trade-offs
Maintain a well-organized list of talks, meetups, or conference contributions (e.g., ICML workshops, NeurIPS posters)
Keep your transcript, resume, and application materials updated and consistent
Infra-focused tips:
Capture SLOs, incident response stories, and infra migrations where you played a key role
Highlight automation and reliability work that directly supported ML or LLM workloads
Document specific technologies and scale (e.g., “Managed Kubernetes clusters supporting 50M daily inference requests”)
Activities such as open-source contributions, mentoring, and speaking engagements demonstrate character and initiative beyond your primary job function.
Fonzi’s profile structure helps candidates present this evidence coherently so letters of recommendation reinforce a clear story rather than adding isolated praise. When a recommender’s letter aligns with your portfolio, hiring teams see a complete picture.
Conclusion
Strong letters of recommendation are essential for AI and infra roles, as portfolios and test scores alone are not enough. The best letters are specific, honest, and tailored to the role or program. Preparation, timing, and clear communication make the process smoother. AI can assist in hiring, but human judgment remains key, and platforms like Fonzi combine technology with expert evaluation.
Identify two to three strong recommenders, organize your materials, and start requests early. Creating a Fonzi profile gives high-signal access to curated Match Days and vetted companies. With the right preparation and strong recommendations, you can navigate the 2026 AI hiring landscape with confidence.
FAQ
How do I ask someone for a letter of recommendation without it being awkward?
What should a strong letter of recommendation include?
How do I write a letter of recommendation for a coworker or employee?
Is there a difference between a letter of recommendation and a reference letter?
How far in advance should I request a letter of recommendation?
