Suggestions
Chiara Piccinotti
Ex-VP of Product & Data Science @ Twitch | Gen AI | Creator Economy | I'm helping a new generation of heart-first, data-savvy leaders unlock their potential.
Chiara Piccinotti is an accomplished professional in the technology industry with a focus on data science, product management, and entrepreneurship. She currently serves as an Executive Board Member at Lesbians Who Tech & Allies, an organization that supports and promotes LGBTQ+ individuals in the tech sector.1
Professional Experience
Chiara's most recent high-profile position was as Vice President of Product & Data Science at Twitch, a leading live-streaming platform.234 This role highlights her expertise in both product management and data science, two critical areas in the tech industry.
Areas of Expertise
- Data Science and AI: Chiara has a strong background in data science and is particularly interested in generative AI and machine learning.23
- Product Management: Her experience as a VP of Product at Twitch demonstrates her skills in this area.234
- Creator Economy: She shows a keen interest in the creator economy, likely stemming from her work at Twitch.234
- Leadership Development: Chiara is passionate about helping develop "heart-first, data-savvy leaders".234
Entrepreneurship and Venture Capital
Earlier in her career, Chiara was involved in entrepreneurship and venture capital. She served as the co-chair of the Entrepreneurship and Venture Capital Group and was the lead organizer for SeedCon 2012, which is Chicago Booth's annual entrepreneurship conference.1
Recent Activities
Chiara is active on LinkedIn, frequently sharing insights on various topics:
- Recruitment and hiring practices in the tech industry5
- Career development skills for data scientists3
- Productivity and time management6
- Customer lifetime value analysis7
- Interview techniques and job application strategies8
Through her posts and activities, Chiara demonstrates a commitment to sharing knowledge and supporting others in the tech industry, particularly in data science and product management roles.
Highlights
Multi-agent collaboration is emerging as a promising way to improve LLM performance. What does this mean for product owners?
Let's start with the basics.
🤖 What's an AI "agent" anyways?
An agent is an AI (usually an LLM) that has been trained (or more likely, prompted) to do a specific task. E.g. to book restaurants, or schedule meetings, or handle customer support requests.
And that works okay... sort of.
There are a few problems: ▶ Single prompts don't handle edge cases well and often deliver unsatisfactory results. ▶ Huge monolithic prompts are a bit like spaghetti code. They're hard to audit and maintain. ▶ Traditional ways of improving performance like fine-tuning and RAGs require data and don't really work as well as everyone had hoped.
🤖🤖🤖 Which is where multi-agent collaboration comes in.
Rather than writing giant, monolithic prompts, AI builders are instead opting to build collections of specialized agents that work in concert.
For example, if I were building a customer service chat bot, I might have the following agents working together: 🤖 A set of "expert" agents that handle different types customer issues 🤖 A "router" agent that determines which "expert" should handle what issue 🤖 An "overseer" agent that monitors the chat for signals that human intervention is needed
Another interesting multi-agent pattern has emerged from the empirical fact that LLMs can improve their responses when prompted by humans to rethink their answer.
Andrew Ng (founder of DeepLearning.AI and Stanford prof) proposed a framework where two AI agents work together to solve coding problems: 🤖 A "coder" agent to come up with the first solution 🤖 A "critic" agent to poke holes and prompt the "coder" agent to improve its solution
What's fascinating about all this is that working with AIs is a lot less like working with software, and a lot more like working with humans.
This is a huge change, and product owners may have to think differently in order to be successful in this space.
Some biggest success factors that I see: 💡 Get used to non-deterministic outcomes 💡 Shift from thinking about architecture to thinking about orchestration 💡 Learn to break down tasks into their primitives 💡 Draw from other disciplines like psychology and organizational design 💡 Build incrementally and manage user expectations 💡 Don't take the human out of the loop too soon 💡 Crank your QA efforts up to 11 💡 Manage your costs thoughtfully (not everything requires GPT 4)
Of course, this is all very new and nascent, and what we believe now may well turn out to be wrong in the future.
Please weigh in! I would love to hear other people's thoughts.
People on social media will tell you they submitted 200 applications to get a job. Don't buy into the "grindset." There is a better way.
If you want to maximize your chances, while preserving your sanity:
✅ Be selective ✅ Get a warm intros wherever you can ✅ Tailor your resume to the job ❌ Don't blindly drop resumes through the company career website
In fact, you should use electronic applications almost as a last resort.
Here's why.
Recruiters get 1000s of applications for each job. Because of this, conversion rates between online application and recruiter screen are often less than 5%.
It's even harder if you have a somewhat non-traditional background, like I had for many years. The likelihood that your resume gets passed over is just very high.
Here's another issue, when you're trying to apply to 100s of jobs, three bad things happen.
1️⃣ You have to cast a wide net, so you end up applying to jobs that are a worse match for you. 2️⃣ It is simply not feasible to tailor your resume to that many jobs, or to properly research that many companies, so you put forward worse applications. 3️⃣ The absolute number of rejections you get increases, which wears on your confidence.
Number 3️⃣ is absolutely crucial.
☝ Confidence is a key ingredient for a successful job search. ☝ It is a scarce resource. ☝ Don't squander it.
You can try to be "zen" about it. You can tell yourself that it's just a numbers game. But the truth is every rejection takes a little out of you.
Too many, and you start losing your confidence.
You get depressed. You start struggling to tell your story with conviction. You end up in a vicious cycle that hurts your chances of getting a job, and can lead to burn-out and depression.
This isn't just diminishing marginal returns. Applying to too many jobs is downright counter-productive.
Don't do that to yourself.
Instead, take it slow. Be selective. Reach out to your network and try to get a warm intro to the recruiter or hiring manager. Really take time to research the company and role, and try to tailor your resume accordingly.
It pays to be intentional here. Good luck! 💕
Thank you for reading!
And if you enjoyed this content, consider reposting to your network, or giving me a follow for more career tips like these.
