Office Hours: I'm head of data and machine learning at Apella Technology and serial founding engineer.Featured

ElphaStaff's profile thumbnail
Thanks so much for joining us @balwinder!Elphas – please ask @balwinder your questions before Friday, August 27th. @balwinder may not have time to answer every questions, so emoji upvote your favorites 🔥👍🏾➕
ngochuynh's profile thumbnail
Hi Balwinder! Thank you for hosting the Office Hour :)My questions are: 1) How do you find the working environment at startups in your experience (what do you like and dislike about it)2) How do you lead teams with diverse personalities and lots of different working styles?3) What does your job look like as a Head of Data and Machine Learning?
balwinder's profile thumbnail
Hello Ngoc,1) At the core of who I am - I'm a dreamer builder - so any role that is taking a product from 0 to 1 or bringing an idea to reality resonates well with me. Startups naturally are a great fit for me. What I dislike about startups generally has more to do with what kind of a startup it is rather than it being a startup. To unpack that - a startup is composed of multiple facets - (a) the problem they are trying to solve, or the product being built (b) how it is being built - or, the engineering practices being followed (c) how invested the startup is in its people who are building the process - the company culture. (d) The ability of the startup to become a sustaining business. When all these come together which generally happens under mature leadership - it is an absolute delight to be part of it. When one or more of these are seriously lacking, the experience becomes distasteful. The one benefit of being part of the founding team is the ability to contribute and influence in all these areas.2) I follow 2 principles - First, tasks are broken up to match the people on the team, not vice versa. Knowing, understanding and respecting team members for who they are is important. Second, the job of an engineering leaders is to provide goals and constraints, be flexible on how the team achieves it. There is a lot of nuance to this - may we can have a separate conversation on that.3) We are in a very fast paced growth period right now, so every quarter looks quite different than the previous one. We do everything-as-code - whether it is managing data, managing the infrastructure to manage data, manage the MLOps pipeline to train our models, and deploy them to production. So most of the time is spent writing code and reviewing pull requests (other team members code). The rest of it is iterating in an agile manner as we learn more about our data, the domain (I am new to Surgery and medTech), and achieving product-market fit. As the AI/ML field is innovating very quickly, striking a balance between adopting the latest, and keeping production systems moving.
ngochuynh's profile thumbnail
Thank you for taking the time to explain in detail!
misna's profile thumbnail
Thanks Balwinder. Always inspiring to see women leaders in tech. What are some practical tips/hacks balancing a demanding career in tech and young children, especially as an immigrant?
balwinder's profile thumbnail
Hello Misna,"It takes a village to raise a child" - so true. So building a network of trustworthy partners is important. The part to not "outsource" is spending time w/ quality time children (which does not equate to doing everything yourself, or doing it perfectly) Most important - believe your work is something that is a positive thing - not a matter to feel guilty about. Children pick up on the underlying emotions more than what you say. Hacks are quite age dependent -Preschool age - having a consistent schedule of pick up and drop off from day care helps.Elementary school age - doing a yearly class presentation in your child's class helps - once kids that age see that their classmates think that his/her mom is so cool, they are very accepting if you can't show up for every field trip. Also, expectation setting "I will chaperone one field trip per year". Junior High/Middle school - give them their space, its a great time to let them fail - they will pick themselves up in high schoolMiddle School/High School - volunteer to be the silent driver who drives them and their friends. Silently listen to the conversations. High School/Under grad - Encourage them to identify their passion, and get behind them to help them realize their dream- that reduces friction about college pathways. Treat them like a friend. Treat them with respect.Working kids - No idea - they have not got off the payroll yet :)Regarding being an immigrant - most of the above would apply to anyone working full time and raising children - the only thing that is different is to balance values from your own culture with the prevalent ones. Again, finding a balance that keeps you confident and not conflicted helps raising kids who are a confident blend of cultures vs. confused kids. I would start w/ finding clarity on what matters to you the most. Kids will follow.
madien's profile thumbnail
Hi @balwinder! Very impressed with your achievements, and thank you for finding the time for Q+A here.My questions are:1. What does a day in your life look like?2. What are your main challenges in ML/Data and how do you plan to conquer them?3. What advice would you give other ML engineers to succeed, who are also daughters, sisters, wives and mothers?Thank you in advance!!
balwinder's profile thumbnail
Hello Medha,1. Well, that depends a lot on the day :) My day is a balance of me time, family time, home maker time, work time & community service time. In all areas - it is a best effort case - I would not say that I am successful. A general day looks like some time for physical and mental exercises, a little bit of meal preparation, looking after family members, working - I love learning and problem solving. We do everything-as-code - whether it is managing data, managing the infrastructure to manage data, manage the MLOps pipeline to train our models, and deploy them to production. So most of the time is spent writing code and reviewing pull requests (other team members code). The rest of it is iterating in an agile manner as we learn more about our data, the domain (I am new to Surgery and medTech), and achieving product-market fit. Finally, I do try to give back to the community by doing some volunteer work. 2. We deal primarily w/ unstructured data - videos and images. 4K videos are huge. Developing a deeper understanding of videos, tradeoffs between different encoding settings, and doing ETL on them is a continuing area of learning and innovation. Tradeoffs between automating pipelines now vs. incurring tech debt to get something out of the door is another area that becomes demanding at times. While it is a great time to be doing ML with the huge amount of innovation in the area of MLOps - it is a noisy space out there - and many times there isn't enough hours in a day to sift through it - to make good judgement calls. How do I plan to conquer them ? By being part of a great team :) When one more than great mind put themselves to a problem and do so focusing on the solution w/ humility - the results are good and iterating fast as we learn more - is the key skill in the toolbox. 3. Keep learning, Treat everyone with respect. Stay fearless. Stay humble. Don't sweat the small stuff.
madien's profile thumbnail
Thank you so much!! Very helpful. If you don't mind me adding a late question (and I understand if you choose to not answer it for the OH are officially over): given that there are newer MLOps products coming to the market every day, which tools/methodologies/techniques have you found success with regarding MLOps?
joclark's profile thumbnail
Hi there thanks for doing OH! My question is given the fight for tech talent at the moment combined with the large amount of capital in the markets, would you recommend an early stage startup buy or build when doing a deep tech/hard tech undertaking? (Meaning the solution almost needs all the pieces to be built to be compelling). Thx
balwinder's profile thumbnail
Hello Jo,Acquiring talent is hard - No doubt about that. It comes down to finding the people who share the passion and vision of what you are doing. I don't think there is a right or wrong answer on - buy vs build. It really comes down to building a defensible business and intermediate milestones to that. Happy to discuss offline.
Sonam's profile thumbnail
Hi there @balwinder ʕ•́ᴥ•̀ʔっWhat's something you enjoy about working in the world of data and machine learning? What's something you find aggravating? What was the biggest change that came after you reinvented yourself? Thank you so much for hosting Office Hours!
balwinder's profile thumbnail
Hello Sonam,The most fascinating thing for me is the ability to solve classes of problems with machine learning models that previous generations of software could not solve.The most aggravating - how much time and effort it takes to manage data. "Reinventing" myself is just another way of making time and effort to learn new techniques to solve new problems. My professional journey has many pauses where I learnt new techniques and moved onto solving a different class of problems. Most of my challenges have not come from "reinventing" myself - they have come when I slowed down on the process of learning, or stopped giving 110% of myself to what I was doing, or let myself get distracted by trivials. As long as I follow my mantra - "Keep learning, Treat everyone with respect. Stay fearless. Stay humble. Don't sweat the small stuff." - life generally treats me well.
Sonam's profile thumbnail
A great mantra indeed! Thank you! (>‿◠)✌
SaloniG's profile thumbnail
Hi Balwinder! Thanks for your time and it's exciting to see how you've been part of building many cool projects including data, IoT and mobile! My questions: - how can you improve yourself from a code monkey to actually knowing how to design and build solutions from scratch?- what is your biggest advice on leading eng teams/ any resources or advice on how to learn?
balwinder's profile thumbnail
Hello Saloni,Reg: "- how can you improve yourself from a code monkey to actually knowing how to design and build solutions from scratch?"Start small - baby steps - everyone gets ideas - you must have had many - try bringing one of them to life. It doesn't have to be perfect. A proof of concept. A small hobby project. A small thing above and beyond what you are expected to do at work. Something for a loved one. Something for a field you care about - art, feeding the hungry, tutoring children, sports - could be anything. Get feedback on it. Maybe others get it, may be they don't. Doesn't matter. Iterate on it. Do some more. Start using what you build. Big companies call it dog-fooding. Soon you will get the hang of it. Hackathons are a great way to build this skill too.- Everyone learns differently. Find what works best for you - Reading books, taking an online course, Having a newsfeed, talking to friends, just building things, reddit, open slack channels - all that matters is that you have a good cadence of learning. I'm alive when I'm learning :)