{"id":1177391,"url":"https://alion.io/job/referred-etworks-oftware-ngineer-lm-da-ui-o-an-umoderu-lmenjinia","title":"Software Engineer - VLM /大規模言語モデル VLMエンジニア","company":{"id":130,"name":"Preferred Networks","domain":"preferred.jp","url":"https://alion.io/company/preferred-networks","size_band":"201-500","is_staffing_agency":false,"is_intermediary":false,"listed_via":null,"ats_vendor":"Talentio","truth_index":null},"role":"Backend","role_family":"Backend","seniority":null,"employment_type":"full_time","work_mode":"hybrid","remote_scope":null,"remote_scope_basis":null,"remote_working_hours":null,"hiring_geo_confidence":"structured","locations":["Tokyo, Japan"],"countries":["JP"],"hiring_countries":[],"hiring_countries_total":0,"salary":null,"salary_estimate":{"min_usd":38000,"max_usd":104000,"period":"year","method":"role_country_seniority_unknown","sample_n":42},"experience_years_min":null,"visa_sponsorship":false,"relocation_package":false,"has_equity":false,"technologies":[{"name":"arXiv","optional":false},{"name":"Computer Vision","optional":false},{"name":"LLM","optional":false},{"name":"Machine Learning","optional":false},{"name":"Multimodal AI","optional":false},{"name":"Post-training","optional":false},{"name":"Pre-training","optional":false},{"name":"Python","optional":false},{"name":"SFT","optional":false},{"name":"Synthetic Data","optional":false},{"name":"VLM","optional":false},{"name":"Amazon CloudWatch","optional":true},{"name":"Amazon EC2","optional":true},{"name":"Amazon EKS","optional":true},{"name":"Amazon S3","optional":true},{"name":"AWS","optional":true},{"name":"DeepSpeed","optional":true},{"name":"FSDP","optional":true},{"name":"IAM","optional":true},{"name":"Kubernetes","optional":true}],"status":"live","first_seen_at":"2026-09-24T12:06:17Z","employer_posted_date":"2026-09-24","last_verified_at":"2026-09-24T16:10:48Z","board_verified":true,"closed_at":null,"days_open":0,"trust":{"level":"ok","repost_count":null,"flags":[],"days_open":0},"description":"Job Description\nPreferred Networks（PFN）では、自社開発の大規模言語モデル（LLM）の技術を中核としたマルチモーダル基盤モデルの開発を進めています。基盤モデルそのものの研究開発に加え、エンターテイメント、科学計算、ロボットなど、PFNの自社事業領域の強化につながる応用開発や、多様な産業への展開を見据えた応用研究にも取り組んでいます。\nVision-Language Model（VLM）チームは、PFNが自社開発しているLLM「PLaMo」を視覚情報と言語を統合的に扱えるように拡張したPLaMo-VLの開発・提供を担っています。モデルアーキテクチャの設計、学習・評価用データの開発、学習、評価から、推論最適化、実環境への導入まで、開発ライフサイクル全体に取り組み、品質、コスト、計算効率などの実用要件を満たすVLMの提供を目指します。必要に応じて既存のVLMや外部APIも活用しますが、本ポジションは、それらを組み合わせるだけのアプリケーション開発やAPI連携を主とする役割ではありません。モデル、データ、評価、推論最適化を含むVLM基盤技術そのものの研究開発・改善が、主な責任範囲です。\n私たちは、研究成果やモデル性能の向上を、PFNのプロダクト、ソリューション、API／プラットフォームとして実用化・提供し、事業課題の解決とPFNの事業成長につなげることを重視しています。事業・プロダクトチームと連携してニーズを技術要件に落とし込み、研究開発から実用化・提供までを一貫して推進します。\n###### 業務例\n- VLMのモデル設計、学習、評価\n- VLMの能力・性能向上に向けた手法の研究開発\n- 社内外の多様なユースケースや要件に対応する、高品質でスケーラブルなデータ生成・整備手法の研究開発\n- 事業・プロダクトチームと連携した、事業ニーズの技術要件への落とし込み、およびプロダクト・ソリューションへの導入・展開\n- 実運用に向けた品質評価、失敗事例の分析、改善\n- 本番環境への導入に向けた推論最適化（エッジデバイスなど、計算資源に制約のある環境への対応を含む）\nPLaMo 2.1-VL Technical Report（[arXiv:2604.19324](https://arxiv.org/abs/2604.19324)）は、モデル・学習設計、合成データ生成・日本語データ整備、ベンチマーク開発、産業現場でのデータ収集・検証まで、こうした業務の幅を具体的に示しています。\n上記は業務の一例であり、実際の担当範囲はこれらに限定されません。\n本ポジションは、特定の応用領域への固定配属を前提としていません。チームが取り組む事業領域や技術課題は、PFNの事業戦略、市場ニーズ、技術の進展に応じて変化する可能性があり、それに伴って担当領域やプロジェクトも変わります。私たちは、一つ以上の領域で深い専門性を発揮しつつ、優先課題の変化に応じて新しい領域を主体的に学び、専門性と貢献範囲を広げられる方を求めています。VLMの研究開発から実用化・提供まで一貫して携わり、その成果を事業課題の解決と事業価値の創出につなげたい方を歓迎します。\n***\nPreferred Networks (PFN) develops multimodal foundation models built on its in-house large language model (LLM) technology. In addition to core research and development on the foundation models themselves, we pursue applied development to strengthen PFN’s business areas, including entertainment, scientific computing, and robotics, as well as applied research aimed at extending these technologies across a wide range of industries.\nThe vision-language model (VLM) team develops and delivers PLaMo-VL, which extends PFN’s in-house LLM, PLaMo, to jointly process visual and linguistic information. We work across the full development lifecycle, from architecture design through data development, training and evaluation, to inference optimization and deployment in real-world environments. Our goal is to deliver VLMs that meet practical requirements for quality, cost, and computational efficiency. While we leverage existing VLMs and APIs where appropriate, this role is not primarily focused on application development or API integration alone. The primary responsibility is the underlying technology itself - models, data, evaluation, and inference optimization.\nWe focus on translating research advances and improvements in model performance into business value by bringing them into PFN’s products, solutions, APIs, and platform offerings. Working with product and business teams, we turn customer needs into technical requirements and drive work end to end from research and development through productization and delivery.\nDepending on your strengths, experience, and the team’s priorities, your responsibilities will span a combination of the following:\n- Model design, training, and evaluation for VLMs\n- Research and development of methods to improve VLM capabilities and performance\n- Research and development of high-quality, scalable methods for generating and curating data for diverse internal and external use cases and requirements\n- Collaboration with product and business teams to translate business needs into technical requirements and deploy VLM capabilities in products and solutions\n- Quality evaluation and improvement for real-world use, including failure analysis and remediation\n- Inference optimization for production deployment, including resource-constrained environments such as edge devices\nOur PLaMo 2.1-VL technical report ([arXiv:2604.19324](https://arxiv.org/abs/2604.19324)) illustrates the breadth of this role: model and training design, synthetic data generation and Japanese data pipelines, benchmark development, and field data collection and validation at an industrial site.\nThese responsibilities are illustrative rather than exhaustive.\nThis role is not tied to any single application domain. The business areas and technical challenges the team works on evolve in response to PFN’s business strategy, market needs, and technological advances. Accordingly, your responsibilities and projects may also change over time. We are looking for people with deep expertise in at least one area who are eager to learn new domains and expand the scope of their contributions as priorities shift. We welcome candidates who want to work end to end, from VLM research and development through productization and delivery.\nQualifications\n本ポジションでは、VLMの研究開発を自律的に推進し、入社後早期から成果を生み出せる実務経験者を想定しています。\n- コンピュータサイエンスおよび数理分野に関する確かな基礎力\n- 業務に関連するアルゴリズムおよび数学（線形代数、確率・統計、最適化など）を理解し、実務上の課題に応用できること\n- 変化への適応力と専門性を広げる意欲\n- 事業・技術上の優先順位の変化に応じて、特定の応用領域や技術手法に限定せず、新しい領域を主体的に学び、担当範囲を柔軟に広げられること\n- 最先端の研究・技術動向を継続的に把握し、関連するコンピュータサイエンスおよび機械学習領域の専門性を主体的に深め、広げられること\n- 研究成果を事業価値につなげる姿勢\n- 事業ニーズや品質、コスト、計算効率、運用上の制約を踏まえ、研究開発の方向性を判断・調整していけること\n- 以下のいずれかを満たすこと\n- 大規模言語モデル（LLM）について、事前学習、事後学習（例：SFTやRL）、評価に関する実務的な知識・実装経験があり、かつ、コンピュータービジョン(例：意味的・空間的理解、特徴点マッチング、カメラ幾何、画像位置合わせなど)に関する研究または実務経験があること\n- VLMまたはその他のマルチモーダルモデルの研究開発経験\n- 視覚情報と言語情報を統合するモデルのアーキテクチャ設計・実装経験\n- 実データまたは合成データを用いた学習・評価を完遂した経験\n- 代表的なVLM手法およびマルチモーダル学習の基本概念に関する理解\n- Pythonによるソフトウェア開発経験（機械学習開発を含む）\n- 再現性・保守性を意識した実装経験（実験管理、コード品質、テスト、デバッグなど）\n- GPU、メモリ、計算量などの計算資源と、レイテンシやスループットなどの実効性能を考慮した実装・改善ができること\n- チームでの課題解決と合意形成\n- 複数のメンバーと協働して技術的な論点を整理し、異なる優先事項を調整して合意を形成したうえで、成果が得られるまで実行できること\n- 英語による業務コミュニケーションに取り組む意思と能力\n- チーム内の会議および日常的なコミュニケーションは英語で行います。翻訳ツールを利用しながらでも、英語でのコミュニケーションに参加できることが必要です。\nThis role is intended for experienced candidates who can independently drive vision-language model (VLM) research and development and contribute from the outset.\n- Strong foundations in computer science and mathematics\n- Solid understanding of the relevant algorithms and mathematics - linear algebra, probability and statistics, optimization - and the ability to apply them to practical problems\n- Adaptability and willingness to broaden one’s expertise\n- Capacity and desire to proactively learn new domains and flexibly broaden one’s scope as business and technical priorities evolve, without being limited to a specific application domain or technical approach\n- Interest in staying current with research and technological advances and proactively deepening and broadening one’s expertise in relevant areas of computer science and machine learning\n- Commitment to translating research into business value\n- Ability to shape and adjust the direction of R&D in response to business needs and real-world constraints related to quality, cost, computational efficiency, and operations\n- One of the following:\n- Practical knowledge of and implementation experience with large language models (LLMs), including pre-training, post-training (e.g., SFT, RL), and evaluation, together with research or industry experience in computer vision (e.g., semantic and spatial understanding, feature matching, camera geometry, image registration)\n- Hands-on research and development experience with VLMs or other multimodal models, including:\n- Experience designing and implementing architectures that integrate vision and language\n- Experience completing training and evaluation using real-world or synthetic data\n- Solid understanding of representative VLM approaches and core multimodal learning concepts\n- Software development experience in Python, including machine learning development\n- Experience building reproducible and maintainable systems, including experiment tracking, code quality practices, testing, and debugging\n- Ability to optimize implementations with consideration for compute resources such as GPUs, memory, and computational cost, as well as practical performance metrics such as latency and throughput\n- Collaboration and technical alignment\n- Ability to collaborate with multiple team members, structure technical discussions, reconcile differing priorities, reach agreement, and follow through to deliver results\n- Ability and willingness to communicate in English at work\n- Team meetings and day-to-day communication are conducted in English. Translation tools may be used, but candidates must be able and willing to participate in English-language communication.\nPreferred Qualifications\n- 大規模学習・分散学習の経験\n- 分散学習フレームワーク（例：FSDP、DeepSpeedなど）を用いた開発・運用経験\n- Kubernetes（K8s）を用いて、データ生成〜学習〜評価までのパイプライン構築・改善経験\n- AWS 等クラウド環境での大規模学習基盤の構築・運用経験（例：EC2/EKS、S3、IAM、CloudWatchなど）\n- データセットまたはベンチマークの設計・構築経験\n- 要件定義、品質管理、アノテーション設計、評価設計（指標・プロトコル）の経験\n- 研究成果を、実際のプロダクト、ソリューションまたは事業成果につなげた経験\n- 実環境や計算資源の制約下におけるモデルの最適化・導入経験（レイテンシ、メモリ、エッジハードウェア、ノイズを含む入力条件など）\n- 3名以上からなる開発チームにおいて、技術面またはプロジェクト面のリーダーシップを発揮し、成果につなげた経験\n- 日本語でのコミュニケーション能力\n- チーム内の会議および日常的なコミュニケーションは主に英語で行いますが、一部の社内文書や情報は日本語のみで提供されています。日本語を理解できる方、または日本語を主体的に学ぶ意思のある方を歓迎します。\n- Experience with large-scale and distributed training\n- Development and operations experience using distributed training frameworks (e.g., FSDP, DeepSpeed)\n- Experience building or improving pipelines spanning data generation, training, and evaluation using Kubernetes (K8s)\n- Experience building or operating large-scale training infrastructure in cloud environments such as AWS, including EC2/EKS, S3, IAM, and CloudWatch\n- Experience designing and building datasets or benchmarks\n- Experience with requirements definition, quality control, annotation design, and the design of evaluation metrics or protocols\n- Experience translating research into deployed products, solutions, or business outcomes\n- Experience optimizing and deploying models under real-world or resource constraints, including latency, memory, edge hardware, and noisy inputs\n- Technical or project leadership experience in a development team of three or more people, with a track record of leading work through to delivery\n- Japanese communication skills\n- Team meetings and day-to-day communication are conducted primarily in English, but some internal documents and information are available only in Japanese. Candidates who can understand Japanese or are willing to learn it proactively are welcome.\nSalary\n経験、業績、能力、貢献に応じて、当社規定により優遇\nExperience, performance, skills, contribution are taken into consideration.\nLocation\n東京都千代田区大手町１-６-１ 大手町ビル / Otemachi Bldg., 1-6-1 Otemachi, Chiyoda-ku, Tokyo, Japan 100-0004\nWork style / 勤務形態\n専門労働型裁量労働制（みなし労働時間：8時間）もしくはフレックス制\nDiscretionary-work (deemed work hours: 8 hours) or Flex-time system\nハイブリッド勤務（オフィス出社と在宅勤務を組み合わせての勤務）\nHybrid work (Combination of working from the office and working from home)\nSalary increase & bonus / 昇給・賞与\n年2回の人事評価及び会社業績に基づいて決定\nBased on the result of a individual performance review (twice a year) and company’s performance\nAllowances / 諸手当\n通勤手当、在宅勤務手当\nCommutation allowances / teleworking allowances\nHolidays / 休日・休暇\n休日：土曜日、日曜日、国民の祝日、国民の休日、年末年始\n当社規定による年次有給休暇制度（入社時26日付与）\n育児休暇、慶弔休暇など\nHoliday: Saturdays and Sundays, public holidays, Year-end and new-year\nAnnual paid leave based on company regulations (26 days granted upon hire)\nParental leave, conguratulation / condolence leave etc.\nWelfare / 福利厚生\n社会保険完備（厚生年金保険、健康保険、雇用保険、労災保険）\n確定拠出年金制度\nラップトップPC購入補助\n定期健康診断実施\nVarious social insurance programs: pension insurance, health insurance, employment insurance, workers’ compensation\nDefined contribution pension\nAllowance for purchasing a laptop PC\nRegular health checks\nEmployment Status / 雇用形態\n正社員（試用期間3ヶ月、本採用と同条件）\nFull-time regular employment (3 months of probation period under the same condition as regular employment)","description_format":"text","description_chars":11465,"description_truncated":false,"requirements":{"experience_years_min":null,"management_years_min":null,"team_size_min":3,"manages_managers":false,"education":null,"security_clearance":false,"languages":[{"language":"Japanese","level":"Upper-Intermediate (B2)","optional":true},{"language":"English","level":"All levels","optional":false}]},"benefits":["Health insurance","Hybrid work","Parental leave"],"hiring_locations":[{"name":"Japan","iso":"JP","kind":"country"}],"hiring_excludes":[],"relocation_offered":false,"industries":["Robotics AI","Foundation Models","AI Chips & Accelerators","AI for Science"],"lifecycle":[{"event":"open","at":"2026-09-24T12:06:17Z"}],"liveness":{"score":86,"band":"hot","label":"Hiring now","p_open":1,"p_active":0.86,"p_room":1,"age_days":0,"expected_fill_days":33,"reasons":["conf:2","win:early"],"computed_at":"2026-09-24T18:42:29Z"},"pay":null,"html_url":"https://alion.io/job/referred-etworks-oftware-ngineer-lm-da-ui-o-an-umoderu-lmenjinia","json_url":"https://alion.io/job/referred-etworks-oftware-ngineer-lm-da-ui-o-an-umoderu-lmenjinia.json","meta":{"generated_at":"2026-09-24T18:42:29Z","cache_seconds":300,"methodology":"https://alion.io/methodology","terms":"https://alion.io/terms","contact":"https://alion.io/contact","api":"https://alion.io/developers"}}