I’m a Senior Machine Learning Engineer who builds production-grade AI systems, with experience spanning large language models, multimodal learning, and agentic AI architectures.

Over the years I’ve worked across NLP, computer vision, speech processing, OCR, and generative models (LLMs, GANs, diffusion) — applying them to real-world systems rather than isolated experiments. My recent work focuses on designing end-to-end AI workflows that combine structured data, documents, images, and contextual signals into reliable, explainable outputs used in production.

I operate at the intersection of ML engineering, system design, and applied AI safety — translating complex models into scalable platforms that can be trusted beyond the lab.

Experience

AI Engineer — additiv (May 2024 – Present · Zurich, Switzerland)

  • Developed Doc-Parser, a core AI platform for automated extraction of structured data from financial and insurance documents — scenario-driven, multilingual, schema-locked, with audit-ready outputs and production APIs.
  • Built a modular agent-based extraction pipeline using agentic AI and LLMs on Azure for document understanding and structured data generation across credit and insurance workflows.
  • Developed ClaimFlow, a multi-agent AI system for end-to-end household and liability insurance claim analysis, covering damage assessment, policy interpretation, and claim consistency checks.
  • Implemented insurance decision workflows using agentic AI and Azure-based LLMs to support claim assessment, broker placement, and claim probability estimation — combining documents, images, and contextual signals into structured, explainable outputs for insurance review teams.

Machine Learning Engineer — Validaitor (Apr 2023 – Mar 2024 · Karlsruhe, Germany)

  • Developed adversarial ML evaluation pipelines to assess and improve model robustness against evasion and poisoning attacks, including FGSM and PGD-based threat scenarios.
  • Built model security assessment workflows including copycat and model-stealing attack simulations, and implemented protection techniques such as watermarking and fingerprinting.
  • Designed LLM validation frameworks to evaluate fairness, bias, and toxicity — enabling systematic risk assessment of models prior to deployment in sensitive and regulated contexts.

Research Assistant — Institut für Parallele und Verteilte Systeme, University of Stuttgart (May 2022 – Oct 2022 · Stuttgart, Germany)

  • Researched spatio-temporal word embeddings, extending Skip-Gram architectures to incorporate temporal and geographic context for time- and region-aware language understanding.
  • Developed Python-based embedding pipelines and evaluation frameworks to analyse temporal semantic shifts and regional language variation across large text corpora.
  • Applied models to downstream NLP tasks including time-sensitive text analysis and regional usage detection, demonstrating improved contextual representation over static embeddings.

Machine Learning Engineer — Quantiphi (Jan 2021 – Oct 2021 · Bengaluru, India)

  • Developed a multimodal emotion recognition system combining text and speech signals into a unified architecture with joint training objectives — F1 score of 0.88 on an internal conversational dataset.
  • Built an unsupervised speaker diarization pipeline using voice activity detection and clustering techniques — diarization error rate (DER) of 17.35% on multi-speaker audio streams.

Machine Learning Engineer — Scanta Inc. (Mar 2019 – Mar 2020 · Gurgaon, India)

  • Took on technical leadership responsibilities, coordinating a small team of ML engineers across the full model lifecycle from experimentation to cloud deployment.
  • Designed and deployed production NLP pipelines on cloud infrastructure for low-latency text processing and language transformation.
  • Developed and integrated data augmentation, language correction, and style transfer models — improving overall NLP system performance by 4.73%.
  • Worked closely with product and leadership stakeholders to translate business requirements into deployable AI solutions and inform product direction.

Machine Learning Engineer — Mobile Programming LLC. (Jul 2018 – Dec 2018 · Gurgaon, India)

  • Developed an attention-based encoder–decoder model for machine translation — BLEU score of 37.13, with improved translation quality for domain-specific text.
  • Built a BiLSTM-CRF model for named entity recognition in pharmacological and medical texts — F1 score of 0.83, integrated into downstream clinical applications.

Education

M.S. Computational Linguistics — University of Stuttgart

B.Tech. Computer Science — Dr. A.P.J. Abdul Kalam Technical University

Skills

Languages & Frameworks: Python, C++, PyTorch, TensorFlow, scikit-learn

LLMs & GenAI: LangChain, HuggingFace Transformers, Azure OpenAI, RAG pipelines, Agentic AI, PEFT / LoRA

Cloud & Infrastructure: Azure (AI Services, Functions, Blob), AWS (Lex, Dialogflow), Docker

ML Domains: NLP, Speech Processing, Computer Vision, OCR, Adversarial ML, Diffusion Models

Other: Knowledge Engineering, Prompt Engineering, Model Evaluation & Safety

Community & Recognition

  • Intel Software Innovator (2019–present)
  • AAAI Reviewer — Safe, Robust and Responsible AI track (2023–present)
  • Member of the Board of Studies at OP Jindal University (2019–present)
  • Author with Packt Publications — Hands-On Deep Learning with TensorFlow 2.0
  • Kaggle 3x Expert — Top 1% of 300k+ users
  • Intel® Edge AI Scholarship (2019)
  • Google Scholarship Recipient (2018)

Interests

Table Tennis, Chess, Financial Markets, Quantum Physics