Sign In

Valliappa Lakshmanan

Data/AI platforms and products

Valliappa Lakshmanan, commonly known as Lak Lakshmanan, is an Operating Executive focused on Data and AI at Silver Lake, a prominent investment firm. He has been in this role since April 2022, working primarily to enhance the value of portfolio companies through innovative data and AI strategies. Lak is based in Bellevue, Washington, and has over 500 connections on LinkedIn.

Professional Background

  • Current Position: Operating Executive -- Data Science at Silver Lake
    • Focuses on leveraging data and AI to drive innovation in portfolio companies.
  • Previous Roles:
    • Director, Data Analytics and AI Solutions at Google (Sept 2019 - Mar 2022): Led a team to develop software solutions for various industries using Google Cloud's analytics and machine learning products.
    • Tech Lead for Big Data and ML Professional Services at Google (Feb 2016 - Sept 2019): Supported partners and customers in utilizing Google Cloud's Big Data and ML offerings.
    • Director of Meteorology (Data Science) at The Climate Corporation (Aug 2014 - Feb 2016): Managed a team of data scientists to create weather-related analytics aimed at improving agricultural yields.
    • Research Scientist at the University of Oklahoma/National Severe Storms Lab (July 1995 - Aug 2014): Developed machine learning algorithms for severe weather phenomena.

Education

  • PhD in Electrical & Computer Engineering, University of Oklahoma (2000-2002)
  • MS in Electrical & Biomedical Engineering, Ohio State University (1993-1995)
  • B.Tech in Electronics & Communications Engineering, Indian Institute of Technology, Madras (1989-1993)

Publications

Lak is also an author with O'Reilly, having written several books on topics such as machine learning design patterns and data governance.

His extensive experience in data science, machine learning, and leadership roles within significant tech companies positions him as a key figure in the intersection of technology and investment strategy at Silver Lake.

Highlights

Jul 31 · linkedin

I tried out two leading agent frameworks -- Autogen and LangGraph. Both are functional and relatively easy to get started with. Between the two, which one should you choose? Considerations include:  

  1. Familiarity of paradigm. Which paradigm (conversational agents or data processing pipelines) are you more comfortable with? I find the data engineering DAG approach a lot more familiar and tractable, but you might find the conversational paradigm more intuitive.  
  2. Level of control and autonomy. In Autogen, because the agents respond to each other, there is quite a bit of heavy handling that has to happen in order to limit such things as an agent responding to itself, or with sequencing prioritization. Some of this is provided by the framework using system prompts that are hidden from you. In LangGraph, the transition rules are under your control, and so the agentic application tends to be easier to build and troubleshoot. However, Autogen can appear more magical, especially when the agents work autonomously. In LangGraph, the amount of autonomy is more limited.  
  3. Strength of ecosystem. At the time of writing, Autogen supports only the OpenAI GPT-4 and higher tool calling capability. LangGraph, because it is based on langchain, supports almost the entire ecosystem of models that support function calling (GPT 3.5, Anthropic, Gemini, Lllama-3, etc.). Also, LangGraph integrates with LangSmith for observability.   Of course, the level of Autogen support for non-OpenAI models and other tooling could improve, LangGraph could add autonomous capabilities, and Autogen could provide you more fine-grained control. The agent space is moving fast!
Jun 14 · linkedin

Business model innovation around Retrieval Augmented Generation from O'Reilly -- they have indexed all the content on Safari and are using it to provide answers grounded in vetted and credible content (their unique sauce). Another innovation here is to credit/compensate authors for the fraction of their work used in a generated answer.

While this is progress, the more well-grounded the summaries look, the more dangerous the inevitable mistakes will be -- irrelevant documents will still lead to misleading summaries, and there will still be hallucinations because of interpolations with the text used to train Llama. At least based on the article, I don't see them employing self-checks or guardrails. I hope those are coming.

https://lnkd.in/gqqHiR2D

Related Questions

What are some of the key projects Lak Lakshmanan has led at Silver Lake?
How has Lak Lakshmanan's work at Google influenced his role at Silver Lake?
What are the main themes in Lak Lakshmanan's books on machine learning and data science?
Can you provide examples of Lak Lakshmanan's contributions to data analytics and AI solutions at Google?
What is Lak Lakshmanan's educational background and how has it shaped his career?
Valliappa Lakshmanan
Valliappa Lakshmanan, photo 1
Valliappa Lakshmanan, photo 2
Add to my network

Location

Bellevue, Washington, United States