N

Nokia

Makes network gear for carriers and data centers

Senior Software Engineer

Full-Time
No salary listed
Senior, Expert
Bachelor's, Master's
India
In Person
AI & Machine LearningApplied Machine Learning

About the job

Requirements
  • Bachelor's or Master's degree in Computer Science, Engineering, or a related field.
  • 8+ years of experience in AI/ML development, including hands-on experience with MLOps tools and serving frameworks, such as Mlflow, n8n, flyte ,TorchServe, etc for model deployment, and Model Context Protocol for Agentic AI Tools.
  • Proficiency in programming languages such as Python, Java, and TypeScript, with a strong understanding of software development principles.
  • Expertise in data handling, particularly SQL, for efficient data management and analysis.
  • Experience with MLOps tools and serving frameworks, such as TensorFlow Serving and TorchServe, for model deployment.
  • Strong background in model engineering and improvement, with a track record of successful model deployment and productionalization.
  • Familiarity with Natural Language Processing (NLP), LLMs and its applications in AI/ML projects.
  • Understanding of CI/CD practices and the ability to develop and maintain pipelines for ML workflows.
  • Excellent problem-solving and analytical skills, with a proactive approach to identifying and resolving technical challenges.
  • Strong communication and collaboration skills, with the ability to work effectively in a dynamic team environment.
Responsibilities
  • Design and implement AI models, leveraging your expertise in AI/ML algorithms and frameworks.
  • Collaborate with stakeholders and engineers to operationalize ML models, ensuring efficient and effective deployment.
  • Develop and maintain CI/CD pipelines for ML workflows, optimizing the model training and evaluation process.
  • Utilize your understanding of RAG, neural networks, and other AI/ML concepts to enhance model performance.
  • Work closely with customers to understand their needs and deliver tailored ML solutions.
  • Stay updated with the latest advancements in AI and MLOps, ensuring our company remains at the forefront of technology.
  • Provide technical guidance and mentorship to junior team members, fostering a culture of knowledge sharing.
  • Ensure data security and privacy compliance in all ML projects, adhering to industry best practices.
  • Collaborate with cross-functional teams to integrate ML models into existing systems and processes.

About the company

Nokia makes the equipment and software that mobile carriers, cloud companies, and businesses use to run networks, from cell-tower radios to fiber links between AI data centers. From 1998 to 2011 it was the world's largest mobile phone maker; it sold that business to Microsoft in 2014, and HMD Global, a separate company, makes Nokia-branded phones under license. Its network unit sells optical systems that send data over fiber, internet routers, and broadband gear, and its mobile unit sells radios and the core software that runs 5G networks. Nokia earns money from equipment and service contracts and from licensing its patents to phone makers and carmakers. It was founded in 1865 as a wood pulp mill and is based in Espoo, Finland.

Company Size

10,001+

Company Stage

IPO

Headquarters

Espoo, Finland

Founded

1865

Get referred to Nokia

See people who can refer or advise you

Simplify Jobs

Simplify's Take

What believers are saying

  • AI builders are a growth engine: AI and cloud customers ordered €2.8 billion from Nokia in the June 2026 quarter.
  • Equity has gained value: Nokia's US shares rose 60% from January to October 2026, while Swedish rival Ericsson's fell 1%.
  • The US is Nokia's top hiring hub: 211 of its 622 open roles in October 2026, versus 19 in Finland.

What critics are saying

  • Cuts are widening: in July 2026, Nokia raised restructuring charges to €800 million from €250 million, targeting Europe and China.
  • Expect office time: only 5 of Nokia's 622 open roles in October 2026 were remote, and 248 were fully on-site.
  • AI is still a small slice: AI and cloud customers made up 9% of Nokia's June 2026 quarter sales.

What makes Nokia unique

  • Nvidia owns 2.9% of Nokia: the companies jointly develop mobile-network software that runs on Nvidia computing hardware.
  • It owns a historic US research lab: Bell Labs, birthplace of the transistor and Unix, lists 10 Nobel Prizes.
  • It manufactures optical chips itself: Nokia owns the San Jose semiconductor plant acquired with US networking firm Infinera.

Help us improve and share your feedback! Did you find this helpful?

Benefits

  • Hybrid Work Options
  • Professional Development Budget

Growth & Insights and Company News

Headcount

6 month growth

3%

1 year growth

5%

2 year growth

5%
Stork
Oct 16th, 2026
Nokia's free Jev rival has a catch.

Nokia's free Jev rival has a catch. A phone-maker's research team has entered the race to make AI decisions faster, cheaper, and more private. But the headline-grabbing promise hides a trade-off that showed up clearly when the models faced real news and satire. Nokia's answer to the AI decision boom. Generative Large Language Models (LLMs) produce text token by token, a creative but often unpredictable process. Decision models, in contrast, return structured outputs: probabilities for typed choices, yes/no answers, or numerical scores. This distinction is critical for applications demanding deterministic, parse-free judgments over freeform text generation. Nokia Applied Research addresses this need with AnyJev, an Apache 2.0 Python library. AnyJev adds a decision layer to open-weight models without requiring fine-tuning or labeled training data. It transforms a generative LLM into a "System One" decision engine, enabling rapid, structured evaluations. AnyJev directly competes with TypeSafe AI's Jev, a hosted, closed-source API. Jev processes approximately 310 tokens per call, costing $0.042 per million input tokens. AnyJev, however, runs locally on your infrastructure with zero per-call cost, leveraging your existing compute. AnyJev supports a range of popular local inference backends: * Hugging Face Transformers * vLLM * SGLang * llama.cpp This flexibility allows developers to integrate robust decision-making capabilities directly into their self-hosted model deployments, bypassing external API dependencies and their associated costs and latencies. The clever trick: read the next token, not a reply. AnyJev's core innovation lies in prompt engineering, not model training. It presents a state (the input text) and a multiple-choice question to an existing LLM, formatting options with distinct labels like "A" or "B." Instead of allowing the model to generate a natural-language answer, AnyJev directly queries the model's next-token probabilities for these specific option labels. This directly retrieves a probability distribution across the choices. This method completely bypasses the risks of malformed text or JSON parsing failures inherent in generative LLM outputs. A model cannot hallucinate or misformat a single character probability. However, a cleanly formatted probability is not inherently correct or trustworthy; it simply reflects the model's raw, uncorrected judgment. To address these inherent biases, AnyJev introduces L0 calibration. L0 tackles position bias by testing every possible ordering of options and averaging the results, ensuring no choice gains an unfair advantage from its placement. It also corrects for label bias by estimating and adjusting the model's prior preference for certain labels, regardless of content. L0 offers two prior estimation methods: batch prior, which analyzes answers across a batch of inputs, and content-free prior, which estimates bias from an input devoid of actual content. This second method is crucial for single-headline judgments, providing a more robust bias correction when batch data is unavailable. L0 significantly improves decision reliability without requiring any labeled training data. 500 headlines later, Jev still wins the big test. Nokia's AnyJev library faced a direct challenge against Jev 1.13 and a raw Qwen 8B model. The benchmark comprised 500 headlines, evenly split between satirical content from The Onion and real news, all processed on an M2 Max MacBook Pro. This test rigorously evaluated accuracy, speed, and confidence calibration. Jev achieved an 80% accuracy rate. Raw Qwen trailed significantly at 68.4%, with AnyJev L0 performing only marginally better at 68.6%. Jev also dominated in speed, completing all 500 evaluations in 51 seconds, while the local Qwen and AnyJev L0 combined took 4 minutes and 36 seconds, even with shared GPU resources. AnyJev L0's core strength emerged in bias correction, eliminating all answer flips when option order changed. Raw Qwen, in contrast, flipped answers on 20 headlines (4%), while Jev saw 18 minor shifts. For further technical details, explore the nokia-applied-research/AnyJev: Turn any LLM into a Jev-style decision model GitHub repository. Confidence metrics revealed a critical flaw in raw Qwen. While 87% of its answers boasted >=99% confidence, it was incorrect on 30% of those highly confident predictions. AnyJev L0 mitigated positional and label biases but did not improve calibration in this specific comparison, leaving Jev as the superior choice for calibrated confidence. Enjoying this? Get one like it in your inbox each morning. one email a day · unsubscribe in two clicks · no third-party tracking Bias correction can fix one test - and break another. Headline trials revealed AnyJev's catch. Qwen, with its strong bias toward satire, often misclassified real news that sounded like The Onion. AnyJev's default L0 calibration, using batch prior, barely corrected Qwen's extreme 0% or 100% probabilities. Applying content-free prior to L0, however, showed a different outcome. This stronger correction helped Qwen recognize real stories that sounded satirical, reducing its lean from 69% satire calls (on a 50/50 split) to a more balanced 67% satire and 66% real news. But this also misclassified some genuine Onion headlines. Correcting a model's label lean helps when that lean is harmful, but can overshoot when the mix of examples changes. The method does not make the underlying model smarter; it merely adjusts its output probabilities. AnyJev is compelling for private, self-hosted decisions and low API costs due to its Apache 2.0 license and local execution. Yet, benchmark AnyJev on your own representative labeled examples before production. The video's results - 80% accuracy for Jev versus 68.6% for AnyJev's L0 - are specific to one model, task, and test setup, not a universal verdict. Frequently asked questions. What is AnyJev? AnyJev is an open-source Python library from Nokia Applied Research that adds decision-style outputs to open-weight language models. How does AnyJev make a decision? It checks the model's next-token probabilities for answer labels such as A or B, then returns probabilities instead of generated text. Did AnyJev outperform Jev in the video's benchmark? No. On 500 headlines, Jev scored 80% accuracy, compared with 68.6% for Qwen with AnyJev's default L0 correction. Is AnyJev free to use? The library is free under the Apache 2.0 license. Running it locally still requires compute resources and a compatible open-weight model. Found this useful? Share it. For builders Want Stork to write one of these about your product? Send Stork a URL. Stork use the product, form a view, and publish what Stork actually think - in 8 languages, labeled Sponsored, with no copy approval on your side. That last part is what makes it worth quoting.

IT Business Today
Oct 16th, 2026
ITBusinessToday weekly News roundup: NVIDIA, panasonic EW, nokia, meltwater & more.

ITBusinessToday weekly News roundup: NVIDIA, panasonic EW, nokia, meltwater & more. Last updated: October 16, 2026 12:41 pm In Tech news this week... NVIDIA has announced a five-year commitment to AI safety in the United Kingdom, highlighting the growing importance of responsible AI development as advanced models and computing infrastructure expand. The commitment reflects ongoing efforts to support AI safety research and strengthen collaboration around the development and deployment of AI technologies. In industrial tech news this week... Panasonic Electric Works and LYNAF are working together to expand smart-lock adoption. The collaboration highlights the growing role of connected access technologies in buildings and facilities, where digital security and convenient access management are becoming increasingly important. In FinTech news this week... Nokia plans to sell its enterprise campus edge business to Aurelius, marking a strategic change in its enterprise connectivity portfolio. The transaction reflects continued restructuring in the technology sector as companies reassess business priorities and ownership of specialized network assets. In MarTech news this week... Meltwater has partnered with Hitachi Solutions East Japan to strengthen market intelligence and corporate communications capabilities. The collaboration brings together media intelligence and technology expertise, reflecting businesses' growing need to monitor public conversations, understand market trends, and make informed communication decisions. In health tech news this week... Zealand Pharma and Generate have partnered to develop AI-designed treatments for metabolic diseases. The collaboration highlights the growing role of generative AI and computational biology in pharmaceutical research, where advanced models are being used to explore potential therapeutic candidates. In GovTech news this week... The BtoB platform INVOICE has partnered with AMAS, highlighting efforts to strengthen digital business processes and transaction management. The collaboration reflects the broader shift toward digital platforms that help organizations streamline administrative workflows and improve business-to-business operations. In HR Tech news this week... KITZ has launched a globally unified human resources platform using SAP SuccessFactors. The initiative is designed to standardize HR processes and employee information management across international operations, supporting a more consistent approach to workforce administration. In IoT news this week... LINE WORKS has begun integration with the BizStack IoT platform, connecting workplace communication with industrial IoT capabilities. The integration highlights the potential to bring operational data and communication tools closer together, helping businesses improve information sharing across field and office teams. Article of the week. AI is becoming an increasingly important part of technology development, from safety research and drug discovery to connected infrastructure and enterprise operations. NVIDIA's five-year UK AI safety commitment reflects the growing focus on responsible AI development, while developments across smart locks, digital communications, and global HR platforms show how organizations are adapting their technologies to changing business needs. Stay tuned for more such updates! ITBT Staffwriter is an in-house contributor at IT Business Today, responsible for delivering timely News and QuickBytes. The author covers the latest developments across IT, tech, and enterprise solutions to keep readers informed with concise and impactful updates. October 16, 2026

iTWire
Oct 16th, 2026
Nokia launches flexible 50G/25G/10GPON solution to accelerate AI-ready fiber broadband.

Nokia launches flexible 50G/25G/10GPON solution to accelerate AI-ready fiber broadband. * Nokia is the first vendor with a 16-port line card solution that supports GPON, 10G, 25G and 50G PON on every port, enabling operators to evolve their networks with greater flexibility and lower risk. * Built on Nokia's Quillion chipset and multi-PON optical modules, the high-density line card unlocks mass-market deployment of fiber access speeds up to 50Gbps. * Nokia's industry-first multi-PON solution enables instant, high-performance, quantum-safe connectivity to meet the demands of the AI era, with general availability planned for 2027. Nokia recently announced a new 50G/25G/10G/GPON line card that gives operators the flexibility to evolve their fiber network at their own pace and be ready for the demands of the AI era. As AI transforms how consumers, businesses and industries use networks, richer content, greater cloud interaction and growing upstream traffic are increasing requirements for capacity, low latency, security and reliability. The solution enables operators to adapt their existing fiber networks as these demands grow. Earlier line cards could limit operators to a narrower set of PON technologies, requiring them to commit capacity in advance or add hardware as market needs changed. Nokia's new multi-PON line card supports multiple generations on every port, allowing operators to introduce higher speeds when and where demand develops. Built on Nokia's in-house Quillion chip and optical PON modules, the new solution supports GPON, 10G, 25G and 50G PON on every port, together with a variety of multi-PON optical modules. This now allows operators to deploy the technology - or combination of technologies - in a flexible way that's best suited to each market, service or business case. Operators can add or change PON generations on the same port and fiber as demand evolves. "The future of broadband is becoming harder to predict. AI is reshaping traffic patterns and creating new expectations for bandwidth, latency, security and reliability. At the same time, operators must make investments in next generation PON that will serve them for years to come. With our new line card, operators do not have to bet everything on a single path. They can deploy the PON technology their market needs now; add additional technologies on the same port and same fiber as the need arises; and easily mix and match PON generations to capture multiple revenue streams," said Maryam Chaudry, SVP & GM, Fixed Networks, Nokia. "As operators prepare their networks for AI-driven applications and higher-speed broadband services, flexibility is becoming just as important as performance. Nokia's ability to support multiple PON generations on every port gives operators a practical, lower-risk path to evolve their fiber networks as demand develops," said Jeff Heynen, vice president, Broadband Access and Home Networking at Dell'Oro Group Powered by Nokia's Quillion chipset and multi-PON optical modules, the line card provides a flexible evolution path from GPON/10G to 25G and 50G PON. For example, operators can deploy 10G + 25G PON today and add 50G PON when demand accelerates. Alternatively, those planning for 10G + 50G PON can easily add 25G PON if market needs develop differently. The solution also enhances network performance, security and resilience: * Up to 4x lower upstream latency for more responsive, data-intensive applications. * Quantum-safe encryption and authentication to safeguard network and user data against emerging threats. * End-to-end resilience without a single point of failure for dependable delivery of mission-critical connectivity services.

APAC News Network
Oct 15th, 2026
Nokia elevates Girish Mehta to Director and Head of NPO.

Nokia elevates Girish Mehta to Director and Head of NPO. Summarize with: New Delhi, Oct 15(APAC Media): Nokia has elevated Girish Mehta to Director and Head of Network Planning and Optimisation (NPO). Mehta has previously worked with Bharti Airtel Foundation and Ericsson India. His experience spans telecommunications and network technology, with his new role focused on network planning and optimisation. India parcel delivery In his new position, Mehta will focus on improving network planning, optimisation and operational efficiency. These functions help telecom operators manage network capacity, improve coverage and maintain service performance as demand for mobile connectivity grows. Network planning helps operators identify capacity requirements and expand coverage, while optimisation focuses on improving the performance of existing networks. Both areas are important for telecom companies seeking to deliver reliable connectivity and manage network resources efficiently. Mehta's experience at Ericsson India adds to his background in telecommunications and network technology. He has also worked with Bharti Airtel Foundation before taking up the leadership position at Nokia. Summarize with:

Light Reading
Oct 15th, 2026
Strategic duality: Navigating HFC and DOCSIS upgrades toward a fiber- and PON-dominated future.

Strategic duality: Navigating HFC and DOCSIS upgrades toward a fiber- and PON-dominated future. An Omdia survey finds that demand for higher speeds, ongoing pressure to reduce operational costs and a changing competitive landscape are driving operators to consider new network approaches. Jim Hodges, Chief Analyst - Cloud and Security, Heavy Reading October 15, 2026 Sponsored by Nokia Hybrid fiber-coaxial (HFC) networks have provided scalable and cost-effective broadband for many years. However, the tremendous demand for higher speeds, ongoing pressure to reduce operational costs and a changing competitive landscape are driving cable operators and other HFC-based network operators to consider new approaches. They are pursuing two primary paths: upgrading their HFC network to increase broadband capabilities and deploying fiber-to-the-premises (FTTP) networks. To gain insight into HFC operators' strategies and priorities, Omdia partnered with Nokia to create a 25-question global web-based survey. The findings are based on responses obtained in May 2026 from 83 employees who worked exclusively for cable operators, municipal service providers or electrical cooperatives. Upgrading HFC networks. Evolving HFC Data over Cable Service Interface Specification (DOCSIS)-based networks through software and hardware updates is not a new strategy, but it continues to be a relevant one - as confirmed in the survey results. While DOCSIS 3.1 and DOCSIS 3.1 Plus/Enhanced are currently the preferred upgrades, the survey results confirmed that DOCSIS 4.0 Extended Spectrum (ESD) at 39% and DOCSIS 4.0 Full Duplex (FDX) at 26% are emerging as strong options for future upgrades. Moreover, as captured in the figure below, these DOCSIS upgrade programs are following an accelerated path; 85% of the operators surveyed forecast they will complete their upgrade programs within the next 24 months. This data, as well as other data from the complete survey report, entitled HFC Evolution and the Rise of PON, reinforces that HFC operators are committed to protecting their investment and upgrading DOCSIS-based networks. When will you complete your DOCSIS upgrade program? Deploying PON. While committed to DOCSIS upgrades in the next 24 months, the survey also verified that HFC operators are looking beyond DOCSIS and taking a new path forward. The technology base that will drive this journey forward is passive optical network (PON). PON-based FTTP networks have emerged as the preferred future path because PON is already well-proven to be reliable, scalable and cost-effective to deploy. Similar to their accelerated DOCSIS upgrade path, HFC operators are adopting an aggressive stance on PON deployment, with 90% planning to deploy PON in brownfield networks and 97% in greenfield networks. In fact, 84% of operators have already started migrating some HFC subscribers to FTTP. Understanding PON variant preferences, with speeds ranging from 10Gbps to 50Gbps, was also a key research focus. With the exception of 50G PON, which is in a pre-commercial trial phase, 10G and 25G solutions are now generally available. Consistent with its maturity and broad availability, most network operators are currently deploying 10G solutions. Of these, XGS-PON is the most common approach. While 25G and 50G deployments will take time to scale, there is already significant interest in deployment. As shown in the table below, more than 9 out of 10 network operators are "currently deploying" (25G = 27%, 50G = 10%), "planning to deploy" (25G = 36%, 50G = 44%), or "considering deploying" (25G = 30%, 50G =39%). This leaves only 7% of operators with "no plans to deploy" either. What are your deployment plans for the following technologies on your FTTP network? The survey research also confirmed that HFC operators believe FTTP/PON will make them stronger on multiple levels. As captured in the figure below, operators believe that FTTP/PON will help them optimize their network, make them more competitive and generate new revenue. Network-centric drivers include future-proofing the network (33%), along with increased reliability/reduced network operational and maintenance costs (32%). The competitive factors include competition from fiber overbuilders (32%) and fixed wireless access (FWA) providers (25%). On the business side, the drivers are to offer new services or address new market segments (29%) and generate higher revenue per subscriber (29%). What are the main factors driving your company's FTTP/PON upgrade? (Select up to three). This blog is sponsored by Nokia. Chief Analyst - Cloud and Security, Heavy Reading Jim leads Heavy Reading's research on the impact of NFV on the control plane and application layers at the core and edge. This includes the evolution path of SIP applications, unified communications (UC), IP Multimedia Subsystem (IMS), session border controllers (SBCs), Diameter signaling controllers (DSCs), policy controllers and WebRTC. Jim is also focused on the network and subscriber impact of Big Data and Analytics. He authors Heavy Reading's NFV and SDN Market Trackers. Other areas of research coverage include Subscriber Data Management (SDM) and fixed-line TDM replacement. Jim joined Heavy Reading from Nortel Networks, where he tracked the VoIP and application server market landscape and was a key contributor to the development of Wireless Intelligent Network (WIN) standards. Additional technical experience was gained with Bell Canada, where he performed IN and SS7 network planning, numbering administration, technical model forecast creation and definition of regulatory-based interconnection models. Jim is based in Ottawa, Canada. Want more Light Reading stories in your Google search results? Join 62,000+ members. Yes it's completely free.