Junior Software Engineer
About the job
- Strong technical fundamentals combined with an intuition for AI model behavior.
- Able to code in Python.
- Able to work independently.
- Design, build, and refine reinforcement learning tasks, where each task is a self-contained software engineering challenge with a prompt, an environment, and an automated grader; own the full lifecycle from idea generation to implementation, running and evaluating frontier models against the tasks, and iterating until the task is rigorous and fair.
- Direct coding agents, evaluate their output, and determine when they are failing in subtle ways.
- Have just graduated or are about to graduate.
- Can code in Python.
- Are confident working independently.
- Are motivated by problems that require both technical skill and creative cleverness.
- No prior machine learning or artificial intelligence experience required.
Today, AI models learn coding from static datasets and toy problems. This team builds the realistic, simulated engineering environments — long-horizon tasks, real tooling, real failure modes — that frontier labs use to train and evaluate the next generation of coding agents.
What you'll work on
- Design and build reinforcement-learning tasks and environments for software-engineering challenges
- Own the full lifecycle: from task concept through evaluation against frontier models
- Evaluate coding-agent outputs and identify subtle failure patterns
- Build Python-based systems, harnesses, and tooling around model training and evals
- Work with Docker/Linux environments that mirror real engineering work
- Own problems end-to-end in a small, fast-moving team
Must-have
- Strong Python skills
- Experience working with the latest AI tools and models
- Exceptional problem-solving ability — strong technical fundamentals and intuition for AI model behavior
- Able to work independently
- In person in San Francisco
Nice to have
- ML or AI experience (e.g., you've trained your own models) — not required
- Strong math/CS background (competitions like USACO/ICPC/olympiads)
- Competed in quiz bowl, debate, robotics
- Impressive personal projects
Why join
- $300K base + equity + bonus — at the junior level
- Your work directly shapes how frontier AI models learn to do real engineering
- Small fast moving team, no bureaucracy — high ownership from day one
- Work at the frontier of RL, agents, and evaluation with the top AI labs as your customers
We're hiring software engineers to build environments and evals for training and evaluating frontier coding agents. Learn more about our interview process and what working here is like. To get a better sense of the work we do, you can check out GBA Eval.
$300K salary · equity · bonus
About the company
Mechanize builds high-fidelity reinforcement learning (RL) environments used by leading AI labs to train models on real-world work tasks. Their simulated work scenarios mimic complex job functions–such as building features, debugging code, and deploying software–allowing AI systems to learn through performance and reward signals rather than static data. Mechanize doesn't train AI models itself or ship consumer products; instead, it focuses on creating the specialized virtual environments that help advance AI capabilities, with a current emphasis on automating software engineering and a long-term vision of enabling AI to handle valuable work across the economy. Backed by prominent tech investors, the company is actively hiring engineers to build these environments.
Company Size
11-50
Company Stage
Seed
Total Funding
N/A
Headquarters
N/A
Founded
N/A
Get referred to Mechanize
See people who can refer or advise you
Simplify's Take
What believers are saying
- Mechanize’s environments target automation of high-value software engineering and office work.
- Strong investor backing enables rapid scaling and development of advanced training data.
- Their focus on robustness addresses key AI limitations like unreliability and long-context performance.
What critics are saying
- Major AI labs could outpace Mechanize by building superior in-house RL environments.
- Technical barriers in automating complex tasks may delay product-market fit for years.
- Dependence on a few large AI labs exposes Mechanize to customer concentration risk.
What makes Mechanize unique
- Mechanize builds simulated digital offices for training AI agents in real-world tasks.
- Their environments support multi-agent collaboration and realistic software tools missing in standard RL setups.
- Mechanize sells RL environments to top AI labs, not consumer-facing automation products.
Help us improve and share your feedback! Did you find this helpful?
Benefits
- Health Insurance
- 401(k) Retirement Plan
- Flexible Work Hours
- Unlimited Paid Time Off
- Paid Vacation
- Hybrid Work Options
- Stock Options
- Company Equity
Company News
Google has a new edge in the AI race. Google has some fresh talent in the AI race. The tech giant has completed its talent deal with San Francisco startup Mechanize Inc, public postings show. Mechanize's co-founder and ex-CEO, Tamay Besiroglu, now works as a research scientist at Google's AI division, DeepMind, according to his LinkedIn profile. Over a dozen other former Mechanize employees now work at Google, per LinkedIn, mostly on the company's 'midtraining' efforts - a key part of building the best chatbot. Google and Besiroglu declined to comment. Google had been in talks for a $1.5 billion-plus deal for Mechanize's tech and talent, Business Insider previously reported. The final terms of the deal weren't disclosed. Mechanize's former chief of staff, Guive Assadi, now describes himself as its CEO, per LinkedIn. Google has been struggling to release an AI model that's widely viewed as competitive lately. It has used talent deals - which are typically structured to avoid antitrust scrutiny compared to fully fledged acquisitions - to gain an edge. For example, Google snapped up AI coding startup Windsurf's talent last year, with former CEO Varun Mohan now leading Google's agentic coding program Antigravity. Mechanize's technology helps tech companies' AI models improve at coding, a challenge Google has struggled with. Mechanize said earlier this year that it raised $9.1 million in a funding round at a $500 million valuation. Besiroglu previously cofounded Epoch AI, which also focused on testing AI models. Have a tip? Contact Hugh via email at [email protected] or Signal at 628-228-1836. Contact Charles via email at [email protected] or on Signal and WhatsApp at 628-282-2811. Contact Katie Roof via email at [email protected] or Signal at @kroof.26
Google quietly closes 1.5 billion dollar talent deal with AI coding startup Mechanize. Google has completed a 1.5 billion dollar-plus talent deal with Mechanize, and the most important part is what it did not buy. Google has found another way to get AI talent without buying the company that built it. According to Business Insider, public postings show that Google completed a deal with San Francisco startup Mechanize, bringing co-founder and former CEO Tamay Besiroglu into Google DeepMind as a research scientist and hiring more than a dozen other former Mechanize employees. That distinction matters. Google did not acquire Mechanize outright. It did not take control of the startup. The final terms have not been disclosed, but Business Insider previously reported that Google had been in talks over a deal worth more than 1.5 billion dollars for Mechanize's technology and talent. Mechanize's former chief of staff, Guive Assadi, now describes himself as the company's CEO, according to LinkedIn details cited by Business Insider. Notice what didn't happen. A full acquisition would have been the obvious story for antitrust lawyers, especially with regulators already watching how the biggest AI companies swallow talent from smaller rivals. This was cleaner on paper: hire the people, license the technology. The corporate shell keeps standing. Mechanize keeps existing. Google gets the part it wanted. This is the workaround. Google used a similar structure with Character.AI in August 2024, hiring co-founders Noam Shazeer and Daniel De Freitas while licensing the company's models. In July 2025, Reuters reported that Google agreed to pay 2.4 billion dollars in license fees for some Windsurf technology and hired CEO Varun Mohan, co-founder Douglas Chen, and some of Windsurf's research and development staff for Google DeepMind's agentic coding work. Google released Gemini 3.8 Flash on September 2, its third Flash-tier model in six weeks, undercutting rivals on price while beating GPT-5.6 Terra and Claude Sonnet 5 on the Terminal-Bench 2.1 coding benchmark. The rapid cadence signals Google is racing to win on cheap, fast models rather than just the biggest ones. - google gemini 3.8 flash model pricing per token - cheap fast ai model releases six weeks cadence What google wanted from Mechanize. Mechanize wasn't a chatbot company. The startup announced itself on April 17, 2025, with Besiroglu, Matthew Barnett, and Ege Erdil saying it would build virtual work environments and benchmarks, plus training data for AI agents. The ambition was not modest: Mechanize said it wanted to help enable the full automation of work. That is a huge claim. The actual product was narrower and more useful to Google: simulated environments and evaluations for long software engineering tasks. In May 2026, Mechanize introduced GBA Eval, a benchmark that gives coding agents 24 hours to build a working Game Boy Advance emulator. Forget autocomplete finishing a line of Python. This is about whether a model can keep track of a messy engineering job for hours. Mechanize said earlier in 2026 that it raised 9.1 million dollars at a 500 million dollar valuation, with backers including Nat Friedman and Daniel Gross, Patrick Collison, Dwarkesh Patel, Sholto Douglas, and Marcus Abramovitch. That's not much money by frontier AI standards. It made the reported Google number impossible to ignore: a company worth 500 million dollars in April was soon attached to talks above 1.5 billion dollars. The price says plenty. Coding is now one of the sharpest fights in AI, because a model that can plan, edit, test, and repair software saves real engineering time. OpenAI has Codex. Anthropic has pushed Claude hard with developers. Cognition built its name on Devin. Google has Gemini and Antigravity, but the Windsurf and Mechanize deals show it was not willing to wait for internal work alone to close the gap. Business Insider reported that the former Mechanize employees at Google are working mostly on midtraining, a stage in model development that helps shape a system's behavior before final post-training. That detail is dry, but it is the point. Google is not just buying a brand name here. It wants people and systems that can help models learn how software work actually unfolds, with false starts, bugs, missing context, and tests that do not pass the first time. The regulatory blind spot. This is the blind spot. Antitrust law is still built around ownership, control, and reportable mergers. These AI talent deals are built to sit just outside that frame. The buyer gets key people and a technology license. The startup continues under new management, which makes the transaction look less like a takeover even when the competitive effect can feel close to one. Paperwork matters too much here. The Washington Post reported in 2024 that Google hired Character.AI's founders and licensed its technology in a deal that stirred antitrust concerns. Reuters reported that the Windsurf arrangement gave Google a non-exclusive license and no controlling stake. Now Mechanize follows the same broad shape, with Google gaining people who know how to train and test coding agents while Mechanize remains an independent company. Marvell Technology's data center revenue jumped 46% year over year to $2.17 billion in fiscal Q2 2027, and the company raised its fiscal 2028 revenue target to roughly $18 billion. CEO Matthew Murphy said custom silicon growth, including an expanded Google deal with an equity warrant attached, is now exceeding the company's own expectations. - Marvell data center chip revenue forecast 2028 - AI infrastructure chip market growth projections That should bother you. If a startup's founder and more than a dozen staff leave for a dominant platform company, the question is not only whether the startup still has articles of incorporation. The question is whether it can still compete in the market it was built to enter. Current merger rules are bad at that question, and Big Tech has noticed. For AI founders, the lesson is blunt: your most likely exit may not look like an IPO or a clean acquisition. It may look like a license fee and a new CEO - with your best team members walking into DeepMind, Meta, Microsoft, Amazon, you name it. Investors can still get paid. Employees can still land good jobs. But if every promising AI tools company ends this way, competition gets thinner even when no one technically buys the company. Google and Besiroglu declined to comment to Business Insider, and the final split between licensing fees, compensation, and other terms has not been disclosed. That missing detail matters. Until regulators decide whether these deals should be judged by their effect rather than their label, the smartest buyers in AI will keep choosing the label that asks the fewest questions. Join the discussion. No replies yet. Start the discussion. Judith Murphy is a financial journalist and market analyst covering AI, technology stocks, and emerging market trends. She has contributed to multiple financial publications and brings a data-driven approach to her coverage of the technology sector and its impact on global markets.
Alphabet (NASDAQ:GOOG) stock price up 1.5% - time to buy? September 11, 2026 Key points. * Alphabet shares rose 1.5% to $335.45 in mid-day trading, though volume was below average. Analysts maintain a consensus "Buy" rating with an average price target of $415.55. * Recent AI expansion is strengthening Alphabet's growth prospects, including the reported $1.5 billion-plus acquisition of Mechanize, broader Gemini Enterprise adoption through Google Cloud partnerships, and Gemini's integration into Avid's creative software. * Key risks include heavy AI infrastructure spending, rising data-center costs and intensifying competition from Microsoft, OpenAI, Anthropic and Meta, which could pressure free cash flow and Alphabet's AI leadership. * Five stocks we like better than Alphabet. Alphabet Inc. (NASDAQ:GOOG - Get Free Report) shares shot up 1.5% during mid-day trading on Friday. The company traded as high as $339.95 and last traded at $335.45. 14,860,364 shares traded hands during mid-day trading, a decline of 28% from the average daily volume of 20,725,635 shares. The stock had previously closed at $330.39. Alphabet news summary. Here are the key news stories impacting Alphabet this week: * Google completed its acquisition of AI-agents startup Mechanize in a deal reportedly valued at more than $1.5 billion. Mechanize's former CEO and more than a dozen employees have joined Google, strengthening Alphabet's ability to develop autonomous AI software, although undisclosed terms limit visibility into the transaction's financial impact. * Google Cloud is broadening Gemini's commercial reach. Gemini Enterprise for Financial Services is entering preview with Deutsche Bank as an early use case, while a new Accenture partnership will train as many as 1,000 engineers to deploy Gemini applications for corporate customers. These initiatives could improve enterprise adoption and help Google Cloud capture more AI-related spending. * Google's Gemini is gaining distribution in professional media workflows through Avid's creative software, increasing its potential to compete with Adobe in video and film editing. Alphabet's TPU business and Google Cloud growth are also being viewed as potential sources of revenue diversification beyond advertising. * Broad technology strength contributed to GOOG's move higher, with Alphabet also lifting major ETFs such as QQQ and VTI. Analysts remain broadly positive, with cited price targets above recent trading levels, but these ETF and analyst signals are not company-specific catalysts. * Alphabet's aggressive AI infrastructure spending continues to pressure free cash flow. Large data-center power requirements may also increase costs, while proposed U.S. rules could make hyperscalers pay the full cost of related grid upgrades. * Competition remains intense from Microsoft, OpenAI, Anthropic and Meta, while ARK Invest's recent sale of Alphabet shares in favor of Meta may reinforce short-term concerns about AI leadership. Wall Street analyst weigh in. GOOG has been the topic of a number of research analyst reports. HSBC set a $515.00 price target on Alphabet in a research note on Friday, August 14th. DZ Bank raised shares of Alphabet to a "strong-buy" rating in a research report on Thursday, July 23rd. Roth Capital upped their price objective on shares of Alphabet from $435.00 to $440.00 and gave the company a "buy" rating in a report on Thursday, July 23rd. Royal Bank Of Canada reissued a "buy" rating and issued a $475.00 price objective on shares of Alphabet in a research report on Thursday, July 23rd. Finally, TD Cowen reaffirmed a "buy" rating and set a $475.00 target price on shares of Alphabet in a report on Thursday, July 23rd. Six equities research analysts have rated the stock with a Strong Buy rating, thirty have given a Buy rating and four have issued a Hold rating to the company. Based on data from MarketBeat.com, the stock has a consensus rating of "Buy" and an average price target of $415.55. Discover more AI infrastructure stocks Trade Financial Derivatives Compare Investment Apps Alphabet trading up 1.5%. The company has a debt-to-equity ratio of 0.16, a quick ratio of 2.64 and a current ratio of 2.72. The business has a fifty day simple moving average of $345.18 and a 200 day simple moving average of $341.69. The stock has a market capitalization of $4.10 trillion, a price-to-earnings ratio of 16.85, a price-to-earnings-growth ratio of 0.95 and a beta of 1.21. Alphabet (NASDAQ:GOOG - Get Free Report) last posted its earnings results on Thursday, July 23rd. The information services provider reported $9.11 earnings per share for the quarter, topping analysts' consensus estimates of $2.87 by $6.24. The company had revenue of $119.80 billion during the quarter, compared to the consensus estimate of $116.53 billion. Alphabet had a return on equity of 51.32% and a net margin of 54.77%.The company's revenue was up 24.2% on a year-over-year basis. During the same quarter last year, the firm earned $2.31 earnings per share. As a group, research analysts expect that Alphabet Inc. will post 20.5 EPS for the current year. Alphabet dividend announcement. The firm also recently declared a quarterly dividend, which will be paid on Monday, September 14th. Stockholders of record on Monday, September 7th will be given a $0.22 dividend. This represents a $0.88 dividend on an annualized basis and a dividend yield of 0.3%. The ex-dividend date is Friday, September 4th. Alphabet's payout ratio is presently 4.42%. Insider buying and selling. In other Alphabet news, Director John Hennessy sold 1,050 shares of Alphabet stock in a transaction on Monday, June 15th. The stock was sold at an average price of $368.63, for a total value of $387,061.50. Following the completion of the transaction, the director directly owned 1,481 shares of the company's stock, valued at $545,941.03. The trade was a 41.49% decrease in their position. The sale was disclosed in a legal filing with the Securities & Exchange Commission, which is available at the SEC website. Also, CAO Marsida Saraci sold 449 shares of the stock in a transaction on Wednesday, July 29th. The stock was sold at an average price of $333.20, for a total value of $149,606.80. Following the sale, the chief accounting officer directly owned 27,386 shares in the company, valued at approximately $9,125,015.20. This trade represents a 1.61% decrease in their ownership of the stock. Additional details regarding this sale are available in the official SEC disclosure. Insiders have sold 451,925 shares of company stock worth $12,371,544 in the last quarter. Insiders own 12.99% of the company's stock. Institutional trading of Alphabet. Several large investors have recently made changes to their positions in GOOG. Castlekeep Investment Advisors LLC bought a new stake in shares of Alphabet during the 2nd quarter worth $29,000. Lifetime Wealth Management P.C. acquired a new position in shares of Alphabet during the fourth quarter worth $38,000. Imprint Wealth LLC bought a new position in shares of Alphabet in the 3rd quarter valued at $31,000. Bard Associates Inc. acquired a new stake in shares of Alphabet during the 4th quarter valued at $41,000. Finally, Timbuktu Capital Management LLC bought a new stake in Alphabet during the 2nd quarter worth $46,000. Institutional investors and hedge funds own 27.26% of the company's stock. Alphabet company profile. Alphabet Inc is a technology holding company best known as the parent company of Google. It was established in 2015 as part of a corporate restructuring intended to separate Google's core internet businesses from a collection of other ventures. Google remains Alphabet's largest business and serves users, advertisers, businesses and developers worldwide. Google's principal products and services include its internet search engine, online advertising platforms, YouTube, Android, Chrome, Google Maps, Google Play and Google Photos. This instant news alert was generated by narrative science technology and financial data from MarketBeat in order to provide readers with the fastest reporting and unbiased coverage. Please send any questions or comments about this story to [email protected]. Before you consider Alphabet, you'll want to hear this. MarketBeat keeps track of Wall Street's top-rated and best performing research analysts and the stocks they recommend to their clients on a daily basis. MarketBeat has identified the five stocks that top analysts are quietly whispering to their clients to buy now before the broader market catches on... and Alphabet wasn't on the list. While Alphabet currently has a Buy rating among analysts, top-rated analysts believe these five stocks are better buys. MarketBeat just released its list of the 7 hottest IPOs expected to hit Wall Street in 2026. See which companies are preparing to go public and why investors are watching closely. Continue following MarketBeat
Google is reportedly in talks for a $1.5 billion-plus deal to license Mechanize and hire its talent just months after the startup’s $9.1 million seed round