Jobs & Careers
Hugging Face’s Latest Small Language Model Adds Reasoning Capabilities
Hugging Face has released SmolLM3, a 3B parameter language model that offers long-context reasoning, multilingual capabilities, and dual-mode inference, making it one of the most competitive small-scale open models to date. The model is available under the Apache 2.0 license.
Trained on 11.2 trillion tokens, SmolLM3 outperforms other models in its class, including Llama-3.2-3B and Qwen2.5-3B, while rivalling larger 4B models such as Gemma3 and Qwen3.
The model supports six languages, including English, French, Spanish, German, Italian, and Portuguese, and can process context lengths of up to 128k tokens, enabled by NoPE and YaRN techniques.
The release includes both a base model and an instruction-tuned model with dual reasoning modes. Users can toggle between different flags to control whether the model generates answers with or without reasoning traces.
Pretraining was conducted over three stages with evolving mixes of web, code, and math datasets. A mid-training phase extended the model’s context length and added general reasoning capabilities, followed by supervised fine-tuning and preference alignment using Anchored Preference Optimisation (APO).
SmolLM3 achieved strong results across 12 benchmarks, ranking high on knowledge and reasoning tasks and demonstrating strong multilingual and coding performance. Instructing and reasoning modes yielded further gains on tasks like LiveCodeBench and AIME 2025.
The full training recipe, including data mixtures, ablations, synthetic data generation, and model alignment steps, has also been made public on its GitHub and Hugging Face pages. This open approach aims to help the research community replicate and build on SmolLM3’s performance.
A few months back, Hugging Face launched SmolLM2, an open-source small language model trained on 11 trillion tokens, including custom datasets for math, code, and instruction-following. It outperforms models like Qwen2.5-1.5B and Llama3.2-1B on several benchmarks, particularly MMLU-Pro, while achieving competitive results on others, like TriviaQA and Natural Questions.
It appears that Hugging Face is focusing on minor but consistent improvements for its small language models.
Jobs & Careers
Bengaluru to Have Its Own ‘Yes San Francisco’ Initiative: Minister Priyank Kharge
Karnataka’s minister of electronics, information technology & biotechnology, Priyank Kharge, announced on July 9 that Bengaluru will soon introduce its own ‘Yes San Francisco’ initiative. This program aims to promote sustainable development in the city while keeping up with the fast-paced technological advancements shaping its future.
Addressing a press gathering alongside KEONICS Chairman Sharath Kumar Bachegowda, Kharge said the move follows his recent visit to the United States for a roadshow, where the team drew inspiration from San Francisco’s collaborative model, co-led by Deloitte and the World Economic Forum. “Bengaluru will be the second city in the world to have such an initiative,” he noted.
The Bengaluru initiative aims to establish a participatory platform that brings together diverse stakeholders, including government bodies, corporations, academia, research institutions, think tanks, startups, and ordinary citizens, to identify and address the city’s pressing urban challenges.
The goal is to co-develop a roadmap for a more inclusive, livable, and future-ready Bengaluru, he said.
The government conducted a consultation involving 45 representatives from MNCs, local companies, Global Capability Centres (GCCs), educational institutions, and think tanks.
The minister said the session surfaced more than 120 potential issues, with sustainability across environmental, social, and economic dimensions emerging as the central theme.
Some of the key areas discussed included solid waste management, greywater management, mobility, public transportation, air quality, green cover, connectivity, and sustainable construction methods.
“Once the problem statements are ascertained, startups, corporations, or incubation centres will receive them and try to find appropriate solutions,” Kharge explained. “Once a stakeholder submits a solution, we will assess its feasibility and offer it as a pilot project. If the pilot succeeds, we will scale it.”
He added that the main goal was to determine how Bengaluru could become a sustainable and livable city in the future.
The initiative is expected to take full shape over the next two months and will mirror the collaborative governance model of ‘Yes San Francisco,’ which has successfully brought together public and private stakeholders to build an inclusive, tech-forward urban policy framework for the Bay Area.
Outer Ring Road Companies Association (ORRCA) had earlier said that IT companies in the city lost over $28 million (INR 225 crore) in a single day on August 30, 2022, as employees were stuck in traffic jams for over five hours.
The industry body said that inadequate infrastructure on ORR has reached a crisis level. Despite 30% of the ORR population returning to working from the office, the collapse of the infrastructure has drawn global concern on the city of Bengaluru’s ability to handle further growth,” Krishna Kumar, general manager of Outer Ring Road Companies Association (ORRCA), had shared.
The state government recently concluded a high-impact 10-day US roadshow in Boston, New York, and San Francisco, aimed at reinforcing the state’s position as a global hub for innovation and technology.
The delegation, led by senior officials from the Department of Electronics, IT & Biotechnology, engaged with over 120 stakeholders, including MNCs, startups, think tanks, and academic institutions.
Kharge said that the recent roadshow secured investments worth ₹5,100 crore (in ESDM and GCC sectors) for Karnataka, which are expected to materialise over the next six months. The investments are projected to generate around 9,700 jobs, he added.
According to a government press release, the initiative included one-on-one meetings, roundtables, and closed-door discussions, focusing on Electronics System Design & Manufacturing (ESDM), biotechnology, and deep-tech innovation. MoUs and Letters of Intent were signed with key players, indicating further investments in the pipeline.
Jobs & Careers
Green Aero Raises $1.6 Million to Build Hydrogen and Defence Aero Engines
Green Aero, an Indian deep-tech startup, has raised $1.6 million in a seed funding round led by pi Ventures, with participation from Antler. The company announced that the funding will support research and development, team expansion, and the creation of an in-house testing facility as the company prepares for commercial pilots in the defence and civil aviation sectors.
It is currently incubated at IIT Delhi and is building propulsion technologies for aerospace, naval, and zero-emission hydrogen aviation applications.
Founded in 2023 by Prithwish Kundu, a former research scientist at the US Department of Energy, Green Aero recently test-fired India’s first hydrogen-based aero engine core, named The Blue Dragon. He was also previously the lead for computational studies and numerical analysis at AgniKul Cosmos, a private launch player in the Indian startup ecosystem.
Indigenous Propulsion Tech
India has long relied on imported turbine engines for aircraft and drones. Green Aero seeks to change this by developing core propulsion systems within the country, targeting defence and sustainable civil aviation markets.
The company is developing aero engines with twice the efficiency of current global models, featuring innovations such as a proprietary turbine, fuel-flexible combustors, swirl-stabilised injectors, and superalloy components created through additive manufacturing.
The startup is also exploring high-thrust supersonic propulsion and plans to launch a small-category commercial engine within the next 12 months. A larger engine platform is under development as part of its long-term roadmap.
Clean Aviation Future
“We are excited to be at the forefront of developing advanced aero engine technologies in India for the world,” said Kundu. “This funding advances our long-term vision to shape a more sustainable future for the transportation industry.”
Shubham Sandeep, managing director at pi Ventures, said, “Green Aero is building world-class aero engines from the ground up in India, with a laser-sharp focus on efficiency and performance. Their bold vision to make India self-reliant in propulsion technology while leading the world in clean aviation deeply resonates with our mission.”
Green Aero’s development strategy combines rigorous R&D, subsystem testing, and innovative design and materials. The company’s goal is to develop lightweight, multi-fuel engines that meet modern aviation demands while minimising environmental impact.
Jobs & Careers
A Beginner’s Guide to AirTable for Data Analysis
Image by Author | Ideogram
Introduction
AirTable is a cloud-based, user-friendly, and AI-driven platform for creating, managing, and sharing databases. It combines the best of Excel spreadsheets with relational database management systems. AirTable offers a freemium subscription model, whereby some limited features can be used for free, making it ideal for smaller projects or beginners, whereas the paid version provides advanced features and a larger amount of computing resources.
This article provides a starting point for anyone interested in AirTable and what it has to offer at a beginner’s level, specifically for data analysis. The article walks you through the process of creating a new AirTable app that incorporates some data and uses it for some basic analysis procedures.
Signing Up and Creating Your First Project
As a cloud-based tool, AirTable does not require downloading a desktop application, but simply accessing its website and signing up. If you have a Google account, for instance, you can use it for a quick sign-up; otherwise, there is the option to register using an email address.
Once signed up, we are ready to create our first project. In AirTable, the concept of base or app is analogous to a project or app — essentially, a container for all the data — so let’s create a new base. If you cannot see at first glance the “create blank app” button, you may have to check for the “Create” button on the bottom-left corner or, alternatively, if there is an “x” icon to be clicked on the top-right corner, click on it and you will be prompted to created a blank app.
You should then see a screen like this:
New base (project) in AirTable
Now it’s time to import some data. AirTable bases consist of one or multiple tables. By default, an empty table named “Table 1” appears. Next to it, there is a tab called “+ Add or import“, which we will click on. In AirTable, there are various options to add data to our project, for instance, from spreadsheets in Google Sheets or Excel, Salesforce, Google Drive, Trello, and many more. We will use one of the simplest approaches: uploading a CSV file, concretely from a URL. To do so, select “CSV file“, and on the left-hand side of the emerging window, choose “Link (URL)“, as shown below:
Uploading CSV data via URL
Copy the following URL to a dataset I made available for you in GitHub, and paste it into the text field that appears. Then click on the right-hand side blue button, and when asked to create a new table or use an existing one, make sure you create a new table. Do not be tempted to use the existing default table called “Table 1”, as that table schema is not compatible with that of the dataset we are importing.
That’s it! You now have a new table populated with the imported data, which contains records of customers in a shopping mall, with the following attributes:
- Customer ID: the numerical identifier of a customer.
- Gender: the customer’s gender, namely male or female.
- Age: the customer’s age expressed as an integer.
- Income: the customer’s annual income in thousands of US dollars ($).
- Spending score: a normalized score ranging between 1 and 100 of the customer’s spending level.
Beginning Data Analysis
In the imported table, all columns are of numerical type, except for “Gender”, which is categorical. In AirTable, a categorical column with one possible value per instance among a predefined set of them is called “Single Select”. You can check or modify the properties of “Gender” or any other field by hovering on the column header, clicking on the v-like icon that appears, and selecting “Edit field”. For this tutorial, we will leave are column types as they are, and proceed to perform some analysis.
Grouping customers by gender: Grouping data records by values of a categorical attribute is as easy as clicking on the “Group” button above the table. Select “Gender” and then “Collapse all” to see aggregated summaries of your data for each gender. By default, you will see the total (sum) of values per attribute and gender, but you can also choose to see the average (or median, min, max, etc.) values of columns like income, spending score, and so on. This can be done as shown in the following screenshot:
Analyzing grouped customers by gender
We can observe that males have, on average, higher income than females, but women spend more than men.
Average income and spending score by gender
To remove a grouping of the data, simply click on the “Group” icon again, then click on the bin icon next to the created grouping to remove it and see your full, ungrouped table again.
Filtering young customers: Next, let’s try to do a filtering. This is an easy and intuitive operation available at the “Filter” icon next to the previously used “Group” icon. In the pop-up dialog, select “+ Add condition”. A filtering condition consists of three elements: a field or column name, an operator, and a value. Examples of conditions are “Age >= 39”, “Spending Score = 10”, “Gender is not Male”, etc. To filter young customers, we will set the condition “Age < 30”. This should filter a total of 55 customers. One interesting thing to do at this point is to combine the filter made with (once more) a grouping by gender, to check whether the findings about income and spending score in males vs. females still apply for young customers. Once you have tried this, filters can be easily removed similarly to groupings.
Using formulae to define an “income class” field: AirTable allows the creation of new columns under many different approaches, formulae being one of them. Simply click the “+” button next to the right-most column in your table to add a new column, and choose “Formula” as the creation method or column type. For instance, we can use the following formula:
IF({Annual Income (k$)} < 40, "Low",
IF({Annual Income (k$)} < 70, "Medium", "High"))
To create a new column called “Income class” whose values (categories) will be defined depending on the value of the annual income column, by following the above formula consisting of two nested conditionals. If you are not familiar with spreadsheet-like formulae syntax, don’t panic, there is a “Create formula with AI” button whereby AirTable’s AI assistant can help build a formula based on your specifications or goal.
Using formulae to create a new column
Using interfaces to visualize your data: Airtable interfaces are used to generate data visualizations. This feature is limited in the free tier, but it is still possible to create simple dashboards with elements like bar charts and pivot tables — that is, cross-column tables that summarize and aggregate the data based on field combinations. To try creating an interface, click on “interfaces” at the top toolbar, and follow the assistant steps. You may end up getting something like this dashboard:
Interface dashboard in AirTable
Note that interfaces are designed to be shareable among teams, e.g., for driving business intelligence processes.
Wrapping Up
This article introduced AirTable, a versatile and user-friendly cloud-based platform for data management and analysis, combining features of spreadsheets and relational databases with AI-powered functions. The guide provided in this article is intended to introduce new users to AirTable and outline some basic functions for data analysis. Needless to say, while they have not been our main focus, AI features provided by the tool are arguably one of the recommended next steps to explore beyond this point.
Iván Palomares Carrascosa is a leader, writer, speaker, and adviser in AI, machine learning, deep learning & LLMs. He trains and guides others in harnessing AI in the real world.
-
Funding & Business1 week ago
Kayak and Expedia race to build AI travel agents that turn social posts into itineraries
-
Jobs & Careers1 week ago
Mumbai-based Perplexity Alternative Has 60k+ Users Without Funding
-
Mergers & Acquisitions1 week ago
Donald Trump suggests US government review subsidies to Elon Musk’s companies
-
Funding & Business1 week ago
Rethinking Venture Capital’s Talent Pipeline
-
Jobs & Careers1 week ago
Why Agentic AI Isn’t Pure Hype (And What Skeptics Aren’t Seeing Yet)
-
Education2 days ago
9 AI Ethics Scenarios (and What School Librarians Would Do)
-
Education2 days ago
Teachers see online learning as critical for workforce readiness in 2025
-
Education3 days ago
Nursery teachers to get £4,500 to work in disadvantaged areas
-
Education4 days ago
How ChatGPT is breaking higher education, explained
-
Education2 days ago
Labour vows to protect Sure Start-type system from any future Reform assault | Children