Karan Keswani

Software Engineer

Delhi, India4 yrs 9 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

Expert in Generative AI and LLMs.
Led AI feature development at Microsoft.
Published research in financial text mining.

Stackforce AI infers this person is a SaaS-focused AI Software Engineer with expertise in machine learning and graphic design.

Contact

Skills

Core Skills

Machine LearningAi DevelopmentGraphic Design

Other Skills

AI ModelsBootstrapCC++Cascading Style Sheets (CSS)Competitive ProgrammingContextual RecommendationsEconomicsFluid FrameworkGPTJavaJavaScriptMongoDBNode.jsPython

About

AI-focused Software Engineer with 4+ years of experience, currently working as an SDE II at Microsoft IDC, driving innovation across Generative AI, LLMs, and intelligent content creation. Led the development of AI-powered features in Microsoft Designer and AI Hub, including GPT-based text-to-design workflows, real-time collaboration using Fluid Framework, and contextual font recommendations. Strong foundation in ML/NLP with hands-on work on BERT, Turing NLG, and DALLE. Published research papers in LDA-based financial text mining. Passionate about building responsible, scalable, and user-centric AI solutions.

Experience

Microsoft

3 roles

Software Development Engineer 2

Promoted

Aug 2023 – Present · 2 yrs 7 mos · Noida, Uttar Pradesh, India

o Implemented Design and Brandkit Sharing capability in the Create Module of AI Hub (an enterprise copilot offering), enabling enterprise users to streamline their design asset management and collaboration.
o Built and deployed a real-time collaboration (RTC) workflow end-to-end using the Fluid Framework, supporting real-time multi-user editing with sub-100ms latency.
o Led the development of a comment feature for design assets, a critical RTC deliverable.
o Developed RAI (Responsible AI) protocols for Designer App mini apps to block malicious content, achieving an F1 score of 0.88 on an internal RAI database, reducing OCV RAI concerns for Microsoft Designer by 90%.
o Designed and implemented a Contextual Font Recommendation Service leveraging distributed architecture to serve personalized font recommendations, ranking 2,000+ fonts in real-time based on user preference and design context. This resulted in a 25% increase in user engagement with recommended fonts on Design Ideas and DFS (Design from Scratch).
o Developed AI-driven content creation features, including Text-to-Carousels/Instagram Stories using GPT (LLM) and Doc-to-Dream Design, converting HTML pages and documents into visually appealing carousels. These features contributed to a 50% boost in user adoption funnel in Designer iOS and Android App.

PythonMachine LearningReal-time CollaborationFluid FrameworkAI DevelopmentResponsible AI+1

Software Engineer

Jun 2021 – Aug 2023 · 2 yrs 2 mos · Noida, Uttar Pradesh, India

o Worked in Microsoft Designer, building features that leverage AI models like LLMs and GANs in graphic designing.
o Implemented and designed critical features like Animation Timeline, Motion in Designer, In-App OCV Feedback, Multiple Text Effects, Text Effects Recommender Engine, Graphic Rendering, Video Controls for Designer App which directly enhanced AI-driven design automation in Microsoft Designer.
o Built a v0 ranker for InspireMe Text Recommendation Engine that relied on STK, Novelty, Variability, and business rules to rank 100+ recipes and finetune itself based on user preferences, with ~1 second latency and 50%+ Kept Rate.
o Implemented InspireMe personalizer that uses data from v0 ranker for human feedback loop-based reinforcement learning, thereby improving kept rate for recommendations by 20%.

Machine LearningGraphic DesignAI ModelsReinforcement Learning