Sign In

Ritendra Datta

Head of Applied AI @ Databricks, Mentor @ MentorCruise

Ritendra Datta is a prominent figure in the field of machine learning, with a rich work history at renowned companies like Google, Facebook, IBM Watson, and Xerox PARC. Currently leading a team of 120 engineers and research scientists, his expertise spans across various domains including ML, infra, and algorithm design for recommender systems, query understanding, text/search analytics, computer vision, bioinformatics, and business analytics. Ritendra's academic journey includes a PhD in the ML/CV domain, which he completed in approximately 4.5 years following his Bachelor's degree. His early fascination with programming since the age of 8 evolved into a passion for statistical modeling and ML during grad school, driven by his love for probability and statistics. Ritendra holds a Doctor of Philosophy (PhD) in Computer Science & Engineering from Penn State University and a Bachelor of Engineering (BE) from the Indian Institute of Engineering Science and Technology (IIEST), Shibpur.

Highlights

Aug 3 · linkedin

In a shocking turn of events, my mentoring actually helped someone! For a few months, I was mentoring someone and always wondered if I was adding any value whatsoever. If she found out I wasn't adding value, I would stop getting the precious $50 a month. Imposter Syndrome of a certain kind.

And then it happened. She got a FAANG company offer for an engineering role and shared the comp details with me. Immediately I sensed a very low-balled offer. She's been out of job for some time and perhaps they were trying to play into the desperation. Slightly annoyed, I wrote a respectful but firm note for her to send to the company, arguing (with data) for why she deserved a $50K higher total comp. She sent it mostly as-is. Next day, the company responded by making her a new offer that was in fact $60K (!) higher, as well as an engineering level bump (!). No new interviews or conversations. Clearly, they really wanted her for the role, and couldn't risk losing her due to what was a tiny sum for the company. She had no other offer, so a bit risky, but it paid off. Literally. If she didn't send that message, nothing would've changed.

With over a 50X return on investment in mentoring, I was relieved. My imposter syndrome paused briefly. Based on one data point, this was definitive proof that my mentoring yields 50X returns. So you should join me. And I should join marketing.

Slightly more seriously: Most companies, for eternity, would continue to offer the lowest possible salary they can get away with, both internally and externally. In a buyer's market, it will only get worse. Remember that when you show loyalty toward a company.

Jul 24 · linkedin

We have created a simple new method to evaluate LLMs. This is joint work with my team member E. Yi Liu and our CTO Matei Zaharia. Evaluating LLM responses at scale is a hard open problem, because there are so many ways to answer the same user questions. And evaluation is the basis for the choices the world would make around GenAI models and systems (e.g. GPT vs. Llama vs. Claude vs. Mixtral). So getting this right is rather foundational.

A scalable way to evaluate one LLM is to instruct a different LLM to evaluate the former's responses, a.k.a. LLM-as-a-Judge. But this works poorly out of the box. A simple but elegant approach, whereby inexpensive per-question 'grading notes' are provided to the judging LLM, agree much more frequently with human judgment compared to a vanilla LLM or a single instruction for all questions. The rates of agreement speak to the effectiveness.

Check out the blog post here: https://lnkd.in/gk7Y6yyi

Also, on a lighter note, with this published note my 'Zaharia number' is now 1 (Erdős number is still 3, I think). For those who don't know, Matei is the creator of Apache Spark.

Related Questions

What are some key lessons Ritendra Datta has learned from his mentees?
What specific challenges does Ritendra Datta help his mentees overcome?
What are some common misconceptions about AI and ML that Ritendra Datta addresses?
How does Ritendra Datta's approach to mentoring differ from traditional methods?
How does Ritendra Datta balance his work and mentorship responsibilities?
Ritendra Datta
Ritendra Datta, photo 1
Ritendra Datta, photo 2
Add to my network

Location

San Francisco Bay Area