Startup Project

Tag: podcast

  • From Meeting Notes to Co-pilot Everywhere: A Product Manager’s Guide to Building Expansive AI Products

    From Meeting Notes to Co-pilot Everywhere: A Product Manager’s Guide to Building Expansive AI Products

    The era of basic AI is over. Product Managers, it’s time to level up. We’ve seen the demos, played with the chatbots, and scratched the surface of what AI can do. But the real game-changer is building AI that proactively assists, optimizes, and anticipates user needs across every aspect of their work. Want to know how to build that kind of next-gen AI product? Listen closely to David Shim, CEO of Read.ai. In a recent Startup Project interview, Shim laid out the roadmap, not just for better meeting summaries, but for a future where AI is a true “co-pilot for everyone, everywhere.” This isn’t just a vision; it’s a $50 million Series B-backed reality. Product Managers, the future of productivity is being built now – are you ready to lead the charge?

    Read.ai, initially known for its AI meeting summarizer, harbors a much grander vision: to be a “co-pilot for everyone, everywhere.” This ambition, backed by a recent $50 million Series B raise, isn’t just about better meeting notes; it’s about fundamentally rethinking how AI can augment human productivity across all facets of work and life. For product managers eager to build truly impactful AI products, Shim’s journey and insights are invaluable.

    Start with the Problem, Not Just the Technology:

    Shim’s story of Read.ai’s inception is a powerful reminder for product managers. It didn’t begin with a fascination with large language models (LLMs) or the latest AI breakthroughs. It started with a personal pain point: the agonizing realization of being stuck in unproductive meetings. “Within two or three minutes of a call, you know if you should be there or not… but now I turned off my camera. I cannot leave this meeting,” Shim recounts.

    This relatable frustration became the seed for Read.ai. For product managers, this underscores a crucial principle: innovation begins with identifying a genuine problem. Don’t get swept away by the hype of new technologies. Instead, deeply understand user needs, frustrations, and inefficiencies. What are the “meetings” – metaphorical or literal – where your users are feeling stuck and unproductive?

    Unlocking Unconventional Data for Deeper Insights:

    Most AI products today heavily leverage text data. Read.ai, however, took a different path, recognizing the untapped potential of video and metadata. Shim’s “aha!” moment came from observing reflections in someone’s glasses during a virtual meeting, sparking the idea to analyze video for sentiment and engagement.

    This highlights a critical lesson for product managers: look beyond the obvious data sources. While text transcripts are valuable, they are just one layer of the story. Consider the rich data exhaust often overlooked – video cues, metadata like speaking speed, interruption patterns, response times to emails and messages. As Shim points out, “large language models don’t pick up” on the crucial reactions and non-verbal cues that humans instinctively understand.

    By incorporating this “reaction layer,” Read.ai’s summaries became materially different and more human-centric, highlighting what truly resonated with participants based on their engagement, not just the words spoken. For product managers, this means thinking creatively about data. What unconventional data sources can you leverage to build richer, more insightful AI experiences?

    Hybrid Intelligence: Marrying Traditional and Modern AI:

    Read.ai’s architecture is not solely reliant on LLMs. In fact, Shim reveals that “90% of our processing was our own proprietary models” last month. They strategically use LLMs for the “last mile” – for generating readable sentences and paragraphs – after their proprietary NLP and computer vision models have already done the heavy lifting of topic identification, sentiment analysis, and metadata extraction.

    This hybrid approach is a powerful strategy for product managers. It emphasizes the importance of building core intellectual property rather than solely relying on wrapping existing foundation models. While LLMs are powerful tools, defensibility often lies in unique data processing, specialized models for specific tasks, and innovative feature combinations.

    Product-Led Growth and Horizontal Market Vision:

    Read.ai’s explosive growth, adding “25,000 to 30,000 net new users every single day without spending a dollar on media,” is a testament to the power of product-led growth (PLG). This PLG engine is fueled by the inherently multiplayer nature of meetings. When one person uses Read.ai in a meeting, everyone present experiences its value, organically driving adoption.

    Furthermore, Read.ai consciously chose a horizontal market approach, resisting the pressure to niche down initially. Shim’s belief that “this is more mainstream… from an engineer at Google leads it to a teacher to an auto mechanic” proved prescient. Their user base spans diverse industries and geographies, highlighting the broad applicability of their co-pilot vision.

    For product managers, this demonstrates the power of designing for virality and considering broad market appeal, especially when building truly transformative products. Sometimes, focusing too narrowly early on can limit your potential impact.

    The Co-pilot Everywhere Vision and the Future of Optimization:

    Read.ai’s evolution from meeting notes to a “co-pilot everywhere” reflects a profound shift in AI’s role in productivity. It’s not just about generating content; it’s about optimization, action, and seamless integration into existing workflows. Shim envisions a future where Read.ai “pushes” insights to tools like Jira, Confluence, Notion, and Salesforce, and also “pulls” data from various sources to provide a unified, intelligent work assistant.

    This vision aligns with the emerging trend of AI agents. However, Shim emphasizes that the real power lies in practical integrations and seamless data flow between different work platforms, rather than just standalone agents. “You want your JIRA to talk with your Notion, to talk with your Microsoft, to talk with your Google, and talk with your Zoom,” he explains.

    For product managers, this means thinking beyond single-feature AI products. The next wave of innovation will be in building interconnected, optimized AI systems that proactively assist users across their entire workflow. It’s about moving from “draft AI” – generating content – to “optimization AI” – driving action and improving outcomes.

    Key Takeaways for Product Managers Building Next-Gen AI Products:

    • Focus on Real Problems: Start with genuine user pain points, not just technological possibilities.
    • Explore Unconventional Data: Look beyond text for richer, more nuanced insights.
    • Embrace Hybrid AI Architectures: Combine proprietary models with LLMs for defensibility and specialization.
    • Design for Product-Led Growth: Leverage inherent network effects and broad market appeal.
    • Vision Beyond Content Generation: Aim for optimization, action, and seamless integration into workflows.
    • Prioritize Value over Hype: Build solutions that deliver tangible ROI and improve user lives.
    • Iterate and Adapt: Constantly learn from user feedback and market dynamics to evolve your product.

    David Shim and Read.ai’s journey offer a compelling blueprint for product managers aiming to build the next generation of AI products. By focusing on genuine user needs, leveraging unconventional data, and envisioning a future of optimized, interconnected AI, product leaders can unlock the true potential of AI to transform the way we work and live.

    Nataraj is a Senior Product Manager at Microsoft Azure and the Author at Startup Project, featuring insights about building the next generation of enterprise technology products & businesses.

    Listen to the latest insights from leaders building the next generation products on Spotify, Apple, Substack and YouTube.

  • How Scispot is redefining modern biotech’s data infrastructure

    How Scispot is redefining modern biotech’s data infrastructure

    Biotech is becoming one of the world’s single biggest generator of data, expected to reach 40 exabytes a year by 2025—outstripping even astronomy’s fabled data deluge. Yet as much as 80 percent of those bytes never make it into an analytics pipeline. Three bottlenecks explain the gap: (1) stubbornly paper-based processes, (2) binary or proprietary instrument file formats that general-purpose integration tools cannot parse, and (3) hand-offs between wet-lab scientists and dry-lab bioinformaticians that break data lineage.

    Verticalization 2.0: Solving for Domain-Specific Friction

    Enter Scispot, a Seattle-based start-up founded in 2021 by brothers Satya and Guru Singh, which positions itself not as an electronic lab notebook or a data warehouse, but as a middleware layer purpose-built for life-science R&D, quality and manufacturing. The strategic insight is subtle and powerful: horizontal cloud platforms already exist, but they optimize for structured, JSON-ready data. Biotech’s heterogeneity demands schema-on-read ingestion and ontology mapping that an AWS or Snowflake cannot supply out of the box.

    Scispot’s architecture borrows liberally from modern data stacks—an unstructured “lake-house” for raw instrument output, metadata extraction via embeddings, and API access to graph databases—but is wrapped in compliance scaffolding (SOC 2, HIPAA, FDA 21 CFR 11) that is prohibitively expensive for labs to build alone. The company is effectively productizing the cost of trust, a move that mirrors how Zipline built FDA-grade logistics in drones or how Databricks turned Apache Spark into audit-ready enterprise software.

    YC’s Real Dividend: Market Signal Discipline

    Although accepted to Y Combinator on the promise of a voice-activated lab assistant, Scispot pivoted within weeks when early interviews revealed that customers valued reliable data plumbing over conversational bells and whistles. This underscores a counter-intuitive lesson from YC alumni: the program’s most enduring value may not be its brand or cheque, but its insistence that founders divorce themselves from their first idea and marry themselves to user-observed pain.

    That discipline paid off. Scispot signed its first customer before writing a line of production code—a pattern consistent with what Harvard Business School’s Thomas Eisenmann calls “lean startup inside a vertical wedge.” By focusing on a tiny subset of users (labs already running AI-driven experiments) but solving 90 percent of their total workflow, the brothers accelerated to profitability in year one and maintained “default alive” status, insulating the firm from the 2024 venture slowdown.

    Why Profitability Matters More in Vertical SaaS

    Horizontal SaaS vendors can afford years of cash-burn while they chase winner-take-all network effects; vertical players rarely enjoy those economies of scale. Instead, their defensibility comes from domain expertise, proprietary integrations and regulatory moats. Profitability becomes a strategic asset: it signals staying power to conservative customers, funds the painstaking addition of each new instrument driver, and reduces dependence on boom-and-bust capital cycles.

    Scispot’s break-even footing has already shaped its product roadmap. Rather than racing to become an all-in-one “Microsoft for Bio” suite, the team is doubling down on an agent-based orchestration engine that lets instrument-specific agents talk to experiment-metadata agents under human supervision. The choice keeps R&D burn modest while reinforcing the middleware thesis: be everywhere, own little, connect all.

    Lessons for Operators and Investors

    1. Treat Unstructured Data as a Feature, Not a Bug. Companies that design for messiness—using vector search, ontologies and schema-on-read—capture value where horizontal rivals stall.
    2. Compliance Is a Product Line. SOC 2 and HIPAA are not check-box exercises; they are sources of price premium and switching cost when woven into the core architecture.
    3. Fundamentals Trump Funding. YC’s internal analysis, echoed by Sizeport’s trajectory, shows no linear correlation between dollars raised and long-term success. Default-alive vertical SaaS firms can wait for strategic rather than survival capital.
    4. Remote Trust-Building Is a Competency. Sizeport’s COVID-era cohort had to master virtual selling and onboarding. As biotech globalizes, that skill set scales better than another flight to Cambridge, MA.

    What Comes Next

    Sizeport’s stated near-term goal is to become the staging warehouse for every experimental data point a lab produces, integrating seamlessly with incumbent ELNs and LIMS. Over a five-year horizon, the company aims to enable customers to mint their own AI-ready knowledge graphs—effectively turning drug-discovery IP into a queryable asset class. If successful, the platform could evolve into the “Databricks of Biotech,” but without owning the data outright.

    Nataraj is a Senior Product Manager at Microsoft Azure and the Author at Startup Project, featuring insights about building the next generation of enterprise technology products & businesses.

    Listen to the latest insights from leaders building the next generation products on Spotify, Apple, Substack and YouTube.

  • Day 5: How to get Insights from YouTube Podcast Video using Open AI & Save Time?

    Day 5: How to get Insights from YouTube Podcast Video using Open AI & Save Time?

    I write a newsletter called Above Average where I talk about the second order insights behind everything that is happening in big tech. If you are in tech and don’t want to be average, subscribe to it.

    A lot of people want podcasts transcribed and read instead of listen to them. We can go one level up and even extract insights from the podcasts as well using Open AI API. Here’s my tweet exchange which provoked this experiment.

    how to create a podcast transcript?

    So here is what we are going to do. And it will slightly different from what I suggested in the tweet. The goal is to pick a YouTube video and get transcription of that video and then using prompt engineering we extract insights, ideas, book quotes & summary etc.,

    To Summarize we achieve our goal in three steps:

    • Step 1: Select a YouTube podcast video
    • Step 2: Transcribe the video
    • Step 3: Get Insights from the transcription

    Step 1: Select a YouTube podcast video

    A recent podcast conversation that broke YouTube was Jeff Bezos on Lex Friedman podcast. So for this exercise I will pick this video.

    Step 2: Transcribe the video

    I used langchain YouTubeAudioLoader along with Open AI’s audio to text model whisper to transcribe the youtube video. As usual you would need your Open AI secret key to use the following script.

    import os
import sys
import openai

from dotenv import load_dotenv, find_dotenv
_ = load_dotenv(find_dotenv()) # read local .env file
openai.api_key  = os.environ['OPENAI_API_KEY']


## youtube video's audio loader - langchain 
from langchain_community.document_loaders.blob_loaders.youtube_audio import YoutubeAudioLoader
from langchain_community.document_loaders.generic import GenericLoader
from langchain_community.document_loaders.parsers import OpenAIWhisperParser #, OpenAIWhisperParserLocal

url="https://www.youtube.com/watch?v=DcWqzZ3I2cY&ab_channel=LexFridman"
save_dir="outputs/youtube/"
loader = GenericLoader(
     YoutubeAudioLoader([url],save_dir),
     OpenAIWhisperParser()
 )
docs = loader.load()
print(docs[0].page_content[0:500])

# Specify the file path where you want to save the text
file_path = "audio-transcript.txt"
try:
    with open(file_path, 'a', encoding='utf-8') as file:
        for doc in docs:
            file.write(doc.page_content)
    print(f'Large text saved to {file_path}')
except FileNotFoundError:
    print(f"Error: Input file '{file_path}' not found.")
except Exception as e:
    print(f"An error occurred: {e}")

    You might see the following error during running this script and I pasted the solution that works in case you are using a Windows system.

    • ERROR: Postprocessing: ffprobe and ffmpeg not found. Please install or provide the path using –ffmpeg-location.

    Running this script will generate the transcript and store it in a text file audio-transcript.txt.

    Step 3: Extract insights from the conversation

    To extract insights, I am using Open AI API and here is the script. The code is loading the transcript text and passing it along with a prompt designed to extract insights, people & books. To get more interesting thigs out of this conversation you can come up with a more interesting prompt. Note that the file name is slightly different because I had to cut the transcript to a short length since my completion query to Open AI API was exceeding my TPM limits.

    import os
import sys
import openai
import shutil
from pprint import pprint

from dotenv import load_dotenv, find_dotenv
_ = load_dotenv(find_dotenv()) # read local .env file
openai.api_key  = os.environ['OPENAI_API_KEY']

client = openai.OpenAI()

file_path = "audio-transcript-copy.txt"
try:
    with open(file_path, 'r', encoding='utf-8') as file:
        long_text = file.read()
    print(f'{file_path} is rad')
except FileNotFoundError:
    print(f"Error: Input file '{file_path}' not found.")
except Exception as e:
    print(f"An error occurred: {e}")

prompt2 = f"""
You will be provided with text deilimited by triple quotes.
The given text is a podcast transcript.

Provide the host and guest name.
Summarize the transcript in to 10 points.

If there are any people referred in the transcript. Extract the people mentioned and list them along with some info about them in the following format
1. Person 1's Name: Person 1's profession or what he or she is known for or the context in which he or she was referred to.
2. Person 2's Name: Person 2's profession or what he or she is known for or the context in which he or she was referred to.
...
2. Person N's Name: Person N's profession or what he or she is known for or the context in which he or she was referred to.
If the transcript doesnt contain refereces to any people then simply write \"No people referred to in the conversation.\"

Extract the books mentioned and list them in the following format.
1. Book 1's Title: Context in which the book was referred to.
2. Book 2's Title: Context in which the book was referred to.
...
N. Book N's Title: Context in which the book was referred to.
If the transcript doesnt contain refereces to any books then simply write \"No books referred to in the conversation.\"

IF you find any inspiration quotoes complie them in to a list.

\"\"\"{long_text}\"\"\"
"""

response = client.chat.completions.create(
  model="gpt-4",
  messages=[
    {
      "role": "user",
      "content": prompt2
    }
  ],
  temperature=0.7,
  #max_tokens=64,
  #top_p=1
)

print(response.choices[0].message.content)

    Here is what the output I got:

    Result: Lex Freidman & Jeff Bezos Podcast Summary

    AI PRODUCT IDEA ALERT 1: Can provide a service to podcasters to generate smart transcripts with insights. This would be a B2B play.

    AI PRODUCT IDEA ALERT 2: Instead of a service to podcast creators, it could be a B2C customers who listen to podcasts and want to get read through podcasts and want to create their own library of insights.

    Expect both these ideas to be used by existing podcast hosting companies like Spotify & launch these ideas as new features. If I was a Product Manager in any such companies I would be pitching them by now.

    That’s it for day 5 of 100 Days of AI.

    Follow me on Twitter, LinkedIn for latest updates on 100 days of AI or bookmark this page.

  • 500x Return from Coinbase Angel Investment

    Liron Shapira is known as the guy who writes twitter threads about web3 use cases including Axie Infinity & Helium.

    But before becoming twitter famous, he is also a founder & angel investor and currently runs a YC backed startup called Relationship Hero.

    The irony in his angel investments is he made life changing money by investing in Coinbase which was a 500x return on his investment.

    Here is a clip of him talking about his Coinbase investment. Listen to get the full details of his investment including original investment and over all returns.

    In the full conversation Liron & Nataraj talked about:

    • Failure with Quixey (50M+ Series C)
    • Relationship Hero (YC Backed)
    • Seed Investment in $COIN (listen for exit multiple)
    • Investing in Russian Bonds
    • Web3 Use cases (cross border payments, helium, NFTs in gaming, Uber on Web3, Identity)

    For Full conversation with Liron check out Episode 24 of Startup Project.

    For interesting outsiders take on business, tech and venture capital subscribe to Outsiders Newsletter. Here is an example post from Outsiders newsletter. Check out how this Patreon makes $220K per month!