Super excited to share the early release of my O’Reilly book “Machine Learning for Robotics” w/
@keerthanpg
🤖
We just released the first two chapters on the platform covering important concepts in robotics, sensors, perception tasks, CNNs and transformers for perception + more!
Excited to share that this summer I’ll be joining
@Tesla
working on computational chemistry and ML for battery materials discovery. I’ll be working on computational and ML research & engineering problems to accelerate comp chemistry, materials discovery and battery development🔋
Building a CPU from scratch right now & it’s fascinating how manipulating a series of bits using specific control logic has led to a foundational piece of tech
Had an awesome week in Texas at the
@Tesla
HQ with my team.
- Got to see the inside of a large-scale cell manufacturing lab
- Learned about processes all the way from R&D, jelly roll creation to cell formation
- Hacked on some new projects in data science/ML for battery chem
I’ve had such an amazing time working at Tesla this summer on the materials cell engineering team 🚀 I got to own projects using physics and ML models for optimizing the battery formation process. As well, got to work on computational chemistry techniques & EE/hardware design!
Starting an informal reading group focused on biology and chemistry! 🧬We’ll dive deeper into the process of protein design, anti-body engineering and cell dynamics. Using various textbooks, plan on covering these topics, let me know if you’re interested!
New project now on
@huggingface
🧪 I built a tool that uses an attention-based graph convolutional neural network technique to accelerate the calculation of reaction energy. This can be used for reactants in a redox reaction for redox flow battery molecules. More in thread👇
Saturday morning activities: delving into
@GoogleDeepMind
’s GNoME paper, exploring the materials project, and learning about inorganic crystal prediction!
I was featured in an article by
@UCBerkeley
where I talk about my journey in startups and research working in ML, robotics, & materials discovery!
Check it out!
Excited to share I’m working with Kleiner Perkins (
@kleinerperkins
) as an investment scout, alongside the talented
@LM_Braswell
. Investing in early stage ideas, especially in ML, robotics, or biotech. I'd love to support if you’re building something, please reach out!
joined
@natfriedman
and other founders over dinner last night. reinforced the idea of "finding a problem you love first and then the best avenue (startup/academia/FRO, etc.) to solve it, rather than the other way around"
Just finished a simulation in Pybullet for Vision-Based Robotic Grasping using reinforcement learning methods such as DDPG. This implementation was widely inspired by this paper: and the Kuka Diverse Object Grasping Environment. Check it out!
I’ll be learning the ins-and-outs of managing largescale data pipelines, building predictive data-driven/ML models to accelerate formulation, discovery & validation of new materials! Even more excited to work with and collaborate with Peter Attia, Matthew Murbach & more :)
Very excited to share my new research paper with
@berkeley_ai
and
@GoogleAI
. We use Google PaLM and other LLMs to extract semantic understanding from scenes that can help robots do mechanical search tasks with 47.1% reduction in number of actions: !
Wouldn’t it be nice if ChatGPT could find your missing keys for you? Our latest research from
@berkeley_ai
+
@GoogleAI
suggests that robots can use large language models (LLMs) to find hidden objects faster. 🧵👇
Quick update: joining The Chan Zuckerberg Biohub to work on multimodal ML imaging for cells/proteins analysis, with applications in viral infections and biochem materials discovery. Excited to collaborate w/
@mattersOfLight
& others. If you’re in SF this summer, let’s meet up!!
How does AlphaFold3 predict the structures of proteins, DNA/RNA, ligands and other biomolecules? I did a deep dive into protein folding, how AlphaFold works and the new model. Here’s what I learned:
(resource in thread)
🚀Excited to share that I’ll be publishing a book, “Machine Learning for Robotics” with
@OReillyMedia
, alongside
@Mascobot
&
@keerthanpg
!
We’re looking for reviewers who have worked w/ ML methods (language, CV, RL/IL) for robotics. If you’re interested, we’d love to connect :)
🧪 Worked on a new project w/ Ankur Gupta and
@merz_garrett
using LLMs to speed up accurate prediction of chemical properties.
To start, we use GPT-3 with GPTChem & GPT2-LoRA to predict the atomization energies of molecules at the G4(MP2) theory level from the QM9 dataset.
Honoured to have
@UCBerkeley
’s Chancellor, Carol Christ write about me in her latest campus letter. The letter was sent out to over 400k+ alumnis and donors :)
Congratulations to
@alishbaimran_
, who has been named in
@TeenVogue
's 21 under 21 of 2022! Alishba, who studies Computer Science at UC Berkeley, is an undergraduate researcher at AUTOLab, a research center run by
@ken_goldberg
.
Just wrapped up an amazing year-end gathering in Toronto for the hard/deep tech community w/
@sana_rfzz
,
@EeshaU
! Over 60+ people jamming out about biotech, ML, robotics and more. Grateful to our sponsors
@BainCapVC
and
@8vc
for their support!🙏
If you're well-versed or interested in RL/IL/robotics and would like to contribute to a piece being published by Springer Nature in the next 1-2 months, please DM me!
Making progress on a minimal prototype for a low-cost hardware self-driving lab system to speed up materials discovery!
The current hardware is <$150 (maker pi pico base, LED, light sensor, pumps). Using bayesian optimization for multi-fidelity and multi-objective optimization.
Team is hiring right now, apply: if any of you are interested in working on these things on the robotics, manufacturing or materials science / battery side.
I’m currently working on a prototype low-cost hardware system using Bayesian optimization algorithms for self-driving labs! 🤖
Interesting to hear about batch and multi-fidelity optimization for improving yield of a reaction or the strength of a material:
Got to try out
@Cruise
’s driverless service in San Francisco last night. So crazy getting to see the ML stack and product that I’m helping build in real-life!
Sharing an open-source model from my research! This is an implementation using an attention-based-CNN-BiLSTM for accurate estimation of lithium-ion batteries state of health (SOH) and remaining useful life (RUL), important aspects of a cell chemistry & it’s aging degree (1/3) 🧵
Excited to share my pre-print “Open Arms: Open-Source Arms, Hands & Control”, with
@hansonrobotics
& in collaboration with
@iiscbangalore
for the ANA Avatar
@xprize
challenge as an effort toward human-AI hybrid telepresence. We present Open Arms 🧵
It was so great getting to chat with
@zebulgar
!
Thanks for the actionable advice on finding your "niche" skillsets/knowledge needed to be a deeptech founder/student, hiring and sharing more about your journey and learnings with
@VardaSpace
🙂
What are best resources for learning fundamentals of jax (automatic differentiation, JIT compilation, custom gradients, etc.) coming from numpy + getting familiar with the structure of jax codebases?
I’ve had an amazing time so far working at
@Cruise
. We are tackling unique challenges that range from hardware (compute, sensors - cameras, radars, acoustics, LiDARs), AI to simulation.
I put together this blog highlighting some learnings & reflections:
Proud to announce that the Brilliant
@alishbaimran_
, along with our Hanson Robotics team has just published a paper about my hands and the platform and hardware that can extend its capabilities. 🤖🦾
Link to Publication:
Fun weekend in DC at the EV unconference! Got to catch up with friends, talk energy/security, and learned a lot. Thanks
@tylercowen
& rest of EV team for putting this together :)
Stanford’s
@hackwithtrees
has been a blast 🚀
Some highlights so far: worked on projects, met + heard talks from
@karpathy
,
@apartovi
, Sajith Wickramasekara (
@benchling
) and it was great catching up with everyone :)
Here’s the updated prototype of our prosthetic arm project using GG-CNNs for grasping poses and learning a control policy using RL, in support of
@SJSU
,
@hansonrobotics
& mentors at the BLINC Lab!Will be sharing a blog with updates on our research & development progress as well:)
The best filter for decision making I’ve found:
- Putting myself in environments with the highest density of ambitious people where I can build and learn the most
- Saying no to commitments that seem interesting but might distract from what I really want to do
1/ Contrary’s mission is to identify the brightest, scrappiest, and most talented builders in the world, and support them for life.
We do this with our community - and for the past 6 years, our Venture Partner (VP) program has been its heart.
Had an amazing time speaking at the
@reworkdl
Deep Learning Summit about research in crystal graph neural networks and autonomous self-driving labs for energy storage and battery materials discovery!
Self-driving robot platform to automatically formulate redox electrolytes for batteries and characterize them without human intervention.
Extensive review paper on software, hardware and applications of SDLs by
@A_Aspuru_Guzik
@M_Seifrid
@felix_s_k
@Stefan_P_Schmid
& others.
@shagun_mm
and I put together this blog to share some of our key takeaways and learnings building a hard tech startup as young first-time founders. We hope this resource helps others who are getting started!
Highlights in thread 👇🏼
17-year-old Hanson Robotics fellow
@alishbaimran_
explains her recent work on improving humanoid manipulation tasks such as grasping using CNNs and a Neural-Symbolic AI approach. Check it out!
New blog post on how we can develop a portable, low-cost, soft robotic glove with Q-learning for sensing+control of the soft actuators for improved motion. This can be used during neuromuscular rehabilitation such as for strokes. Let me know what you think!
Grateful to be awarded the Eclipse Grant to further my work in comp mat sci/bio! Huge fan of
@mollyfmielke
’s work supporting young people. Recommended applying if you are looking for non-dilutive funding for your research or side projects!
Exciting to have our reflective paper in
@digital_rsc
(thanks for the effort
@kmjablonka
)!
Check it out to learn more about how LLMs are being applied to chemistry and materials:
Going to be sending out an update on what I’ve been up to past few months; research, ML chem/mat sci work, my book, and much more!
You can sign up here if you want to receive the update + future updates directly:
A glimpse at what I’ll talk about:
Sat down to chat with Berkeley’s new college (
@BerkeleyDataSci
) about my O’Reilly book and current research in RL & rep. learning at
@berkeley_ai
!
We have some exciting new chapters dropping soon on language models, simulation, and RL!
Check out the interview highlights:
Looking forward to speaking at the
@Gartner_inc
IOCS Conference in Las Vegas next week where I’ll be speaking about my work in energy storage & ML!
I’ll be in Las Vegas from Dec 5-8th. If you’re around then, would love to catch up or chat :)
Had an amazing time organizing a learning retreat with
@daviderady
and
@MukundhMurthy
! Spent 2 days in the woods learning, bonding, and sharing ideas:
Published a pre-print paper on some of our work
@hansonrobotics
with the AHAM team
@xprize
. We outline a framework using machine perception, convolutional neural networks, and symbolic AI for logical control and affordance indexing of robotic arms:
Never read so many research papers in a day as I am now that I’m writing a book 😅
Learning how to quickly extract knowledge at different levels of abstractions is underrated. And knowing when to switch between a higher (system) & lower level (architecture) in your explanations.
Simulated control of a soft bio-inspired robot gripper using Q-learning to train the control policy for a simple manipulation task. Done with
@SofaFramework
using SRT by Harvard Biodesign lab! Will also be sharing a blog explaining how this works soon. Check out this short demo!
Have been reflecting a lot on what it means to unapologetically be yourself:
- Acts as a “filter” to attract people with aligned values
- You get better at taking risks and facing failure/rejection
- Builds confidence
Wrote a piece sharing my thoughts:
Had a very fulfilling day today, realized I'm the most satisfied when I'm:
- learning a lot & pushing myself
- working on an exciting mission/project
- surrounded by amazing people, supportive and deep conversations with friends
Meet
@AlishbaImran_
, UC Berkeley student entrepreneur studying CS, book author of "Machine Learning for Robotics", and previous startup founder of Voltx. She recently joined
@Tesla
to continue her work on machine learning. Follow her for updates on her book and awesome work!
I’m so honoured to be one of
@TeenVogue
’s 21 under 21 this year! It’s very inspiring to to be included on this list with other young people doing amazing work in their fields :)
Exploring faster pathways to synthesize and characterize materials for discoveries in batteries, semi-conductors, etc.
Just configured the hardware and controller system for my color-mixing self-driving lab, a low-cost version of a costly multi-step chemical discovery SDL!
Had a great time speaking to executives from leading companies about sustainability in their companies at
@Gartner_inc
’s global conference! I touched on my work in ML, automation, & important areas in accelerating renewable energy storage, materials discovery/recycling.
Excited to share my pre-print paper publication with
@SJSU
!We've developed a novel design & 3D printed material to reduce the cost of prosthetics from $10k to $700 on average, equipped with a closed-loop vision-based manipulation technique to improve use.
Been getting lots of questions on how to get the O-1 (extraordinary ability/achievement visa) and the process.
Put together a short article highlighting some of my learnings and tips:
Hope this is helpful, DMs open if anyone has any questions!
Excited to be launching this blog series with
@olivercameron
! We'll be doing research deep dives and covering exciting advancements in ML, robotics, materials science, computational science, and much more :)
Learn more in this thread 👇🏼
Really excited to share our first deep-dive later this week. The ramifications of it are insane.
Subscribe here to be ready:
Proud to partner on this with
@alishbaimran_
!
Hosting an end of the year get together in Toronto (w/
@sana_rfzz
) for young people in hard/deep tech (biotech, hardware, ml, etc). There will be good food, project demos and lots of time to mingle!
RSVP here:
Interact is a community and fellowship full of some of the most authentic, inspiring, and helpful friends I’ve ever met. Def recommend applying if you’re a young technologist!
Just had the
@beondeck
#odf
8 kickoff call and it was most of the most well-put-together virtual experiences I’ve ever had! There was great culture building and energy in the call + I met really interesting people. My thoughts about communities from this experience in thread!
My talk at
@reworkdl
Deep Learning Summit was just published!
I talk about graphs representation of materials with physics, applications of GNNs for materials property prediction, synthesis of materials, and autonomous self-driving labs.
Check it out:
Hi Cal friends! Holding another builder session this Sunday at 10am. Come out to hack on your projects alongside other founders, students, & researchers.
Last event had a great turnout + amazing projects in batteries, microdrones, ML & more. Sign up here:
Good reminder I’ve been thinking about! Resonate a lot with this framework of compartmentalizing your time & thinking about problems you wanna work on for 4-5 yrs of your life. This is especially helpful when thinking about exploring vs. spending time building depth.
Pattern I've noticed in very impactful people: Spend ~1 year exploring broadly, ~4 years relentless focus executing on the most interesting direction, repeat.
I’ve been working with
@hansonrobotics
+
@singularity_net
to create SophiaDAO. This organization is designed to nurture the development of advanced human-like robots such as
@RealSophiaRobot
& AGI research. Learn more in this blog I wrote:. More in thread👇🏼
Just shared a quick update on what I've been up to for the past few months! I touch upon:
- my internship at
@Cruise
- research
- publishing a book with
@OReillyMedia
- travels and much more!
Check out the update here:
Learnt & grew the most this summer with other amazing founders as a fellow at the Delta Fellowship by
@HumandotCapital
. Really grateful for all the tactical advice/learnings which I applied directly to grow my startup.Thank you to
@barisakis
&
@TheEmma_Kirst
for this opportunity!
Spent the day at the
@czbiohub
in SF learning about cell segmentation techniques & generating embeddings for tomograms. Learned so much & about exciting projects. Thanks
@mattersOfLight
,
@Adrian_Jacobo
& others for having me!
People often underestimate the compounding effects of learning best practices in a field and immersing in problem-solving before building. Having depth can be crucial to building something substantial in a field.
It was great to chat with Dave Patterson today. He's someone I've looked up to, a turing award winner & inventor in computer architecture!
We spoke about many things, including his research but here are some key learnings that I thought I'd share + may be helpful for others: 👇🏼
Check out this new paper featuring our project:
Great compilation of how LLMs are being applied to chemistry/matsci for predictive modelling, tools, & knowledge extraction.
Thanks
@kmjablonka
,
@BenBlaiszik
,
@andrewwhite01
for putting this together!
Can’t wait to be part of the first cohort of Commit and to meet/learn from amazing people like Sarah and the other fellows. Highly recc applying it you’re a student, new grad or researcher building in AI/ML!
1/ 👇Apply to Commit,
@w_conviction
's community program for young AI builders, kicking off with an IRL hackathon sponsored by
@OpenAI
@huggingface
@basetenco
on 6/1-6/4. Meet co-conspirators, hack, and win $$, LLM/GPU credits and glory. Travel covered
Hacked away on FastTrack this weekend, an application to provide female founders with funding opportunities through connecting them with investors, technical advisors and allowing them to build in a safe space with other founders. Sign up for beta here:
Understanding the atomic-level structure and dynamics of macromolecules, such as proteins, can be useful for designing new materials or drugs. I wrote a small piece on this problem, cryo-EM and how deep generative modelling is used:
I’m helping organize builder sessions (Whiteboards and Waffles) with
@contrarycapital
at Cal! Come out this Sunday to hack on passion projects alongside other ambitious Cal founders, researchers, scientists, and builders. RSVP here to attend:
📄Excited to share our paper which was published in
@ASMEdotorg
! We study the design of affordable prostheses and scalable learning with off-policy algorithms for grasping with a 78% grasp success rate on previously unseen objects.
@SJSU
@SJSUEngineering
Just gave a talk at TruffleCon 2019, with over 700+ Blockchain industry leaders in Seattle. So insanely grateful for everyone at
@trufflesuite
,
@Microsoft
&
@theksociety
that helped make this possible :) - Offical talk & pictures will be our soon!!