Stethoscope & Algorithms. Medical AI. Building https://nitter.cf/t.co/eShNnX3MIp.

pale blue dot
Joined April 2022
This is now available. It's the most comprehensive collection of Nigerian treatment guidelines and covers 270 medical conditions. Can be used to build RAG pipelines or to validate your health AI application. You can access it here:
I now have about 270 medical conditions cleaned and clinicaly validated based on the Nigerian Standard Treatment Guidelines. I believe they can be used for benchmarking, model fine-tuning, etc. Will be making it publicly available soon.
37
90
8
314
36,183
Chisom Rutherford retweeted
Today we’re introducing Gemini 4 Argon. It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit.
967
3,165
1,764
29,722
3,571,679
On it already!
NEW Viewpoint: The clinician–artificial intelligence scientist: a proposed career pathway in medicine. Read it here: thelancet.com/journals/landi…
2
2
19
1,043
Lol this is evidence that your v1 doesn't have to be perfect. Even a chatbot for the US Government is not perfect.
Well ain't that something
1
1
3
319
I've been using GPT 6.1 Sol as soon as it launched, and the last time I felt any improvements was with 5.6 Sol. Maybe I'm not giving it difficult tasks, or maybe the models are so good now that newer models only show marginal improvement.
1
3
268
Chisom Rutherford retweeted
doctor by training, eczema patient by experience, web designer by necessity I redesigned Moya’s website myself today if you live with a chronic skin condition, would this make you want to try the app? be brutally honest 👇
5
1
14
799
Chisom Rutherford retweeted
1
3
17
This could significantly cut down the time spent on developing evals.
11
618
Chisom Rutherford retweeted
hiring clinical vs non-clinical PMs as more clinical people get into tech, there are two schools of thought when it comes to building clinical product teams: - hire a clinical PM who can own the product and work directly with engineers - hire a traditional PM who works with engineers and have them interface closely with clinicians on your team or with the clinical users of the product both are viable paths with pros and cons. i’m of course biased (as an MD doing product work) but understand why some companies may want to go down the path of hiring a traditional PM let’s walk through pros and cons of both and when you should hire each one clinical PMs when to hire - you’re working on a product where the end user is a clinician - you’re building a clinical AI tool that needs clinical evals - the connection between clinical mindset and product needs to be very aligned pros - as clinicians they can be the end user and understand the user journey - you save a hire and lots of communication and coordination that is needed when the product person and the clinical person are the same person. things can go from this person to the engineering team - user feedback is much faster because if the end users are clinicians they feel much more comfortable talking to another clinician and feel seen and heard cons - it’s hard to find people who fit this category and are truly good at both product and medicine - finding someone who is deeply technical on top of this or at least can get in the weeds with eng can be difficult - they may not have depth in both medicine and product (i.e they are good at both but maybe not incredible at either) non-clinical PMs when to hire - your product touches healthcare but not necessarily clinicians directly - you have a clinical team already that can easily pair with product / eng or you are hiring for roles where you want embedded clinical people in product / eng teams pros - they are much easier to find - many times you can find someone who has worked for many years at large tech companies (e.g. Google, Meta) and knows how a product machine works. they can hit the ground running and build this for your start up - because the pool of applicants is larger, you can find people across the spectrum of skills (e.g. more design oriented PMs vs more technical PMs) cons - they may not understand healthcare and medicine as much and may need to rely on clinical counterparts which can slow things down - initial product ideas and MVPs require clinician review - more cooks in the kitchen and more roadblocks till something can get to eng - you need to hire both product staff and clinical staff or outsource the clinical work and knowledge that is needed to build the product you may find that you eventually need both types of people at your company as you grow but thinking through if your first product hire should be clinical vs non-clinical is an important step regardless, if you’re building a clinical product you will need to factor in how to get clinician input one way or another
9
3
26
2,699
It's why rigorous evals are important. In a RAG system, you're not only evaluating the outputs, but also the retrieval, and that's the only way to know what is working and what is not.
There are lots of RAG techniques and I don't think as an engineer you can know all. I think what matters is knowing which RAG technique(s) is best suitable for your project/problem
3
2
1
19
1,001
I have been able to make cool professional motion graphics animations with Claude Opus 5.5 Great stuff... If you are not using ai you are cheating yourself
1
2
2
81
Chisom Rutherford retweeted
There's nothing better than a Lustré weekend. Voila 🪄
2
2
64
We should measure technological progress by adoption, not by capabilities. The models have become highly capable, but have we been able to adopt them to solve real problems?
6
5
41
712
Chisom Rutherford retweeted
Your statement contains a painful truth but at the same time, it is a false dichotomy. Even the truths are being addressed: 1. You commented on data centers: data centers indeed require a massive amount of electricity, which Nigeria's national grid can't support. However, it is also noted that Nigeria has nearly 25 data centers, the second highest in Africa. All are Tier III and Tier IV, with 99.99% uptime. It's mainly run by companies like Equinox and even MTN. They operate independent power plants that are insulated from the national grid. 2. Software cannot create mass employment alone: yes, this is true. Software alone cannot save Nigeria. The issue, of course, lies in the labour density of this field. Today, financial services and telecom are major drivers of GDP growth. However, despite that, the tier 1 banks combined employ only a little over 30,000 people. The main sectors that can create mass employment are manufacturing, agriculture, and trade. However, your statement falls apart this way: 1. There's a wrong assumption in Nigeria that simply copying the traditional method of industry (like China) can lift millions out of poverty single-handedly. This isn't correct in the 21st century. I think this is what individuals who cite China and Vietnam miss. You don't need to choose digitalization vs. industrialization. These industries now run together. Put simply, the application of digitalization in industry- what I call tech-driven industrialization is it. Digital payments, financing, etc are the financial rails that facilitate industrialization. 2. It seems we can't draw a direct line between this digitalization and actually making life better for people. It sounds as though digitalization is preventing industrialization. If I ask how this industrialization should happen, you will probably say the government. Even before the startup/digilization boom, has the Nigerian government really done anything viable in this regard? Imagine if the founder of oPay decided that he should wait for 24/7 electricity before they build. Or if Moniepoint, Bamboo, or even our industrialist Dangote decided to wait for that. Like it or not, we have benefited. If these things weren't built, the Nigerian government still wouldn't have done anything. 3. Let's go back to electricity. The truth is, to a mogul looking to build an industry in Nigeria, electricity isn't at the top of the list of headaches. It isn't the major problem. All the major industries rely on independent power generation. It's cheaper and more effective. Even if Nigeria has uniform Band A electricity today, those industries aren't going to port. Again, at that scale, electricity isn't a problem. To that individual, government policy and regulatory friction are a bigger problem, not electricity. @0xkitng, there was a post you made that covered this; I will appreciate it if you may, Sir. 4. What does owning our AI actually mean? Well, it is known that this isn't the time to build frontier models. Europe doesn't have one, and basically the USA and China are the main competition in this space. Hence, even developed nations are struggling. It is estimated that 80% of data center construction plans in the US were discontinued this year due to power. However, it would be a mistake if Africa were to use industrialization as an excuse not to play in this space. Technology is immersive, disruptive, and incremental. 5. A Notion that Nigeria isn't already industrializing: it mightn't be as obvious, but Nigeria IS industrializing, funny enough. Local industries are doing amazing things. @Rxbremen was making a thread on various companies doing amazing things already; they aren't waiting for 24/7 electricity, good roads, or a steady internet connection "Owning AI" does not mean Nigeria must build a $10 billion supercomputer cluster to train a trillion-parameter model from scratch. We're talking about: i. Local data centers: @NITDANigeria is at the helm of data sovereignty; the CBN has mandated financial institutions in Nigeria to utilize local data centers. MTN is going big into this space. @Rxbremen, there was a post you made about data centers in Nigeria ... that included a chart, will appreciate. ii. Sovereign Data & Context: African datasets, dialects, and medical realities shouldn't be erased by models trained exclusively on Western data. We don't need to wait for 24/7 electricity for that one. Unless you realize that after achieving electricity, all that is gone. iii. Domain-Specific Adaptation: fine-tuning open-source models across various industries and building on foundational models to help across various industries like agriculture, healthcare, and fraud detection. Should these problems persist when we have the tools that can make life better for people? iv. Economic defense: What happens if these countries shut Nigeria out? If Dangote waited for 24/7 electricity for his refinery, how exactly would we have navigated this energy crisis engulfing the globe? Mind you, Nigeria hasn't built a functional one in over 30 years. Digitalization was not even a thing, and we didn't industrialize. The truth is that the rules of industrialization have changed. Okay, purely manufacturing: how do you want to outcompete the Chinese? In summary, I agree that industrialization plays a critical role in Nigeria's development. However, industrialization must be done right, paired with digitalization, to actually move the needle. One mustn't come before the other; both MUST move together. This is 2026, not the 80s.
As a Third World country, we can’t keep skipping industrialization and jumping straight into digitalization. We cannot be talking about owning our own AI when we don’t have the electricity day to day activities less to talk about a data center
9
8
1
28
3,086
Good work!
Millions of people turn to AI to navigate difficult relationships, work through stress, or support someone they care about. We’re introducing MentalHealthBench, an open benchmark developed with mental health experts to measure how well AI responds in these moments. 🧵
10
379
Same. This is truly a different era.
I have shipped more stuff in the last 1 year than in my entire career.
1
5
369
We just published a framework for using local AI to triage TB patients in low-resource areas with few or no radiologists.
26
81
6
374
7,293
I strongly believe in cheap, adaptable technological solutions for our healthcare problems in Nigeria/Africa. This is what this work reflects, and is also why we're building housejob ng.
We just published a framework for using local AI to triage TB patients in low-resource areas with few or no radiologists.
5
10
58
1,057