Portfolio

Senior AI Engineer. I'm building and running the systems that turn messy web data into production-grade pipelines.

Skills and Interests

  • Programming Languages: Python, TypeScript, SQL, NoSQL, Golang, JavaScript
  • Frameworks: Pytorch, Flask, Django, React, Angular, EmberJS, ExpressJS, Semantic UI, D3.js, Next.js, shadcn, Tailwind CSS
  • Developer Tools/Environment: Linux, Git, Docker, Kubernetes, ElasticSearch, PostgreSQL, Redis, Heroku, Helm, ArgoCD, GCP
  • Machine Learning: NLP, Deep learning, DP SGD, Classification, Segmentation, Computer Vision, Multimedia Retrieval
  • Leadership: Technical ownership of platform initiatives, client-facing engineering, incident management & RCAs, mentoring through code review, cross-team communication
  • Natural Languages: English(Bilingual), Hindi(Native), French(Elementary), German(Elementary)
  • Interests:Hiking, Biking, Mobile Photography, DJ House Music

Professional Experience

Kadoa

Kadoa is an AI platform that turns unstructured web data into production-ready data pipelines for hedge funds, hiring platforms, and enterprises.

Senior AI Engineer

March 2026 - Present | Basel/Zurich, Switzerland

I lead forward-deployed engineering at Kadoa: the operational core of the platform, where it meets the customer. I own the function and ship every day. My approach: document everything, keep things simple, automate the rest.

Support & Operations Automation

  • Support lifecycle: centralized ticket creation, dedup, incident clustering, and auto-resolution behind one endpoint. Around 95% of resolutions need no one from Kadoa.
  • Agent operations: CI-guarded agent instructions and self-service guidance so AI agents repair workflows without engineers.
  • Audit logging: every human and agent action across operations is logged.

Data Platform & Reliability

  • Workflow chunking: designed the architecture with chunk declaration and execution APIs, atomic aggregation, bounded retries, and fan-out caps. Scaled extractions from a low-thousands ceiling to millions of records.
  • Operations dashboard: replaced per-card polling with one batched feed and live updates over WebSocket, and cut the interface surface roughly in half.
  • Public API & SDKs: removed legacy navigation modes, protected internal state, made generated IDs deterministic and schema-driven.

Clients & Team

  • Enterprise accounts: direct technical owner for requirements, incidents, and RCAs across scraping, anti-bot infrastructure, and data quality.
  • Reviews: 300+ PRs in five months across a team of 8+ engineers.
  • Knowledge base: built and run the engineering docs platform, architecture decision records, and team playbooks.

Tech Stack

  • Languages & Frameworks: TypeScript, Node.js, Python, React, Next.js
  • Data & Infra: PostgreSQL, DuckDB, ScyllaDB, Redis, BullMQ, GCS/MinIO, Kubernetes, Terraform, GCP
  • AI: LLM extraction pipelines, agentic workflow automation, MCP, prompt evaluation

AI/ML Engineer (Working Student)

October 2024 - February 2026 | Basel/Zurich, Switzerland

Joined part-time alongside my MS. Each project ended with me owning a larger slice of the platform.

  • Email service: replaced two legacy email services with one central API with React templates, sanitization, and rate limiting. Migrated all services to it.
  • Events & notifications platform: event-driven architecture with email, Slack, webhook, and websocket channels, handling 10K+ events a day with retries and dead-letter queues.
  • LLM extraction accuracy: image preprocessing and a benchmarking harness for financial PDFs, reaching 95% accuracy.
  • Cost & usage: usage ledger and aggregation pipeline over 1M+ rows with per-customer cost APIs.
  • Delivery infrastructure: image tagging pipeline, ArgoCD hooks, Cypress in CI, migration tooling.
  • Client workflows: high-volume production extraction in the final months, the on-ramp to the role above.

European Organization for Nuclear Research (CERN)

Software Engineer II

October 2021 - August 2023 | Geneva, Switzerland

Worked on CERN Analysis Preservation (CAP), a service for researchers to preserve and document components of physics analyses, ensuring outputs are preserved, findable, and accessible by collaborators in the future.

Platform Development

  • CAP Client CLI: Designed and implemented a command-line interface to interact with CAP APIs, enabling 10,000+ researchers to use the service efficiently.
  • ADL Parser: Developed a parser for Analysis Description Language (ADL), facilitating particle physics data analysis and onboarding 3,000+ researchers.
  • Search Functionality: Enhanced search capabilities using Elasticsearch's painless scripting, implementing customizable sorting options to improve result relevance.
  • Code Quality: Introduced automated workflows to enforce PEP 8 standards, improving developer experience.

DevOps & Infrastructure

  • Health Check System: Implemented a health status page for CAP APIs in Python, deployed on Kubernetes with cron and sidecar containers for real-time monitoring.
  • CI/CD Optimization: Reduced end-to-end testing time by 50% by leveraging pre-built Docker images in continuous integration workflows.
  • Infrastructure Migration: Migrated from Elasticsearch 5 to OpenSearch with zero downtime, ensuring uninterrupted service and improved infrastructure resilience.

Documentation

  • Documented the CAP architecture to streamline onboarding for new developers.
  • Created a detailed CAP CLI usage guide to support the research community of 10,000+ users.

Challenges & Solutions

  1. Onboarding New Experiments:

    • Challenge: Integrating varied experiments from the Large Hadron Collider (LHC) into CAP, each with unique workflows and data requirements.
    • Solution: Designed personalized JSON schemas for each experiment, seamlessly aligning with existing collaboration workflows.

  2. Infrastructure Resilience:

    • Challenge: Ensuring minimal downtime and robust performance for CAP's infrastructure.
    • Solution: Created QA and test instances using Argo CD and comprehensive end-to-end tests for new features.

  3. Data Search & Retrieval:

    • Challenge: Improving the accuracy and relevance of search results within CAP.
    • Solution: Leveraged Elasticsearch painless scripting to provide customizable sorting options in JSON schemas.

  4. Access Control:

    • Challenge: Implementing granular access permissions to manage sensitive analysis information.
    • Solution: Designed custom permission levels within the schema, ensuring secure and appropriate data accessibility.

Tech Stack

  • Languages & Frameworks: Python, JavaScript, Flask, React
  • DevOps Tools: Docker, Kubernetes, Redis, Celery, PostgreSQL, Argo CD, Helm
  • Search & Monitoring: Elasticsearch, OpenSearch, Kibana

Leadership

  • Led backend development by collaborating closely with the product manager and project team lead, ensuring alignment with physicist requirements and technical goals.

Software Engineer I

August 2020 - Sep 2021 | Geneva, Switzerland

Worked on REANA, a reproducible analysis platform enabling scientists to run containerized data analysis pipelines on remote compute clouds, and on CERN Open Data, a platform simplifying user access to 3 PB of particle physics data.

REANA Platform Development

  • REANA Client CLI: Built a command-line interface for REANA APIs, supporting 10,000+ researchers in running reproducible analyses.
  • Job Resource Management: Developed a feature to customize job memory limits at the API and CLI levels, improving resource allocation for the Kubernetes backend.
  • Data Persistence: Ensured durability of RabbitMQ message queues, preventing data loss during pod failures.

CERN Open Data

  • Search Optimization: Improved search functionality by introducing nested facets in Elasticsearch, enabling better usability and exploring 3 PB of particle physics data.

Challenges & Solutions

  1. CLI Development

    • Challenge: Addressing security concerns and ensuring input parameters were validated correctly in the REANA CLI.
    • Solution: Implemented validation and pre-check mechanisms on the client side, reducing the risk of errors and ensuring smooth server-side operation.

  2. Job Management

    • Challenge: Memory exhaustion and inefficient resource allocation during computational tasks.
    • Solution: Enabled customizable job memory limits for the Kubernetes backend, optimizing performance across varied workflows.

  3. Data Persistence

    • Challenge: RabbitMQ queues lost data during pod failures, impacting workflow execution.
    • Solution: Made RabbitMQ queues durable, ensuring data persistence and platform reliability.

  4. Search Optimization

    • Challenge: CERN Open Data searches lacked usability for exploring large datasets.
    • Solution: Introduced nested facets in Elasticsearch, greatly enhancing the search experience and usability for researchers.

Tech Stack

  • Languages & Frameworks: Python, Flask, Click, JavaScript, React
  • DevOps Tools: Docker, Kubernetes, RabbitMQ
  • Search & Monitoring: Elasticsearch, Kibana

Leadership

  • Took ownership of defining project requirements and delivering the REANA CLI as per the needs of researchers, ensuring the platform was user-friendly and efficient for thousands of users worldwide.

Working Student / Part Time

Weblate

Software Engineer II

September 2023 - May 2024 | Remote

Contributed part-time to Weblate, a web-based continuous localization platform supporting over 2,500 libre software projects and companies in 165+ countries.

Platform Development

  • Screenshot Synchronization: Introduced a feature to automatically sync screenshots from the git repository, ensuring up-to-date visual references for translators and reducing manual workload for maintainers.
  • Webhook Development: Designed and implemented webhooks to trigger, sync, and log events across projects, enabling seamless integration and improving configuration synchronization at project-wide and component levels.
  • Add-on Management: Enhanced configuration by introducing project-wide and site-wide add-ons, simplifying management across large-scale localization projects.

Challenges & Solutions

  1. Synchronization Issues

    • Challenge: Screenshots in the repository often fell out of sync with the code, leading to outdated visual references for translators.
    • Solution: Implemented an automated syncing mechanism that aligned screenshots with the latest version.

  2. Configuration Complexity

    • Challenge: Maintaining add-ons across numerous components was time-consuming and prone to inconsistencies.
    • Solution: Developed features to enable centralized add-on management at the project and site levels, reducing complexity and ensuring consistency.

  3. System Integration Complexity

    • Challenge: Integrating external systems via the REST API required syncing with individual translations, which was inefficient.
    • Solution: Introduced webhooks to streamline and automate integration tasks, minimizing errors and saving maintainers' time.

Tech Stack

  • Languages & Frameworks: Python, Django,, Bootstrap
  • Tools & Technologies: REST APIs, Git, Celery, Redis, PostgreSQL

Leadership

As an advocate of the open-source ecosystem, I took ownership of impactful features, directly enhancing user workflows and reducing maintenance burdens. My contributions reflect a commitment to improving usability and efficiency for thousands of Weblate users.

Internships

European Organization for Nuclear Research (CERN)

January 2020 - June 2020 | Geneva, Switzerland

  • I worked on the Indico, used daily at CERN to manage more than 600,000+ events of different complexities and 200+ meeting and conference rooms.
  • The United Nations also uses Indico to handle more than 180000+ participants in conferences happening worldwide and thousands of visitors.
  • My work was focused on implementing category-scoped groups and improvements in registration/abstract statistics in Indico.
  • Tech stack: Python, Flask, JavaScript, React, PostgreSQL

Premiere Digital Services

September 2019 - December 2019 | Remote

  • I worked on Storefront, which is the world's most powerful SaaS platform for managing global content for 20+ platforms, including Netflix, Amazon Prime, Youtube, Disney, and many more.
  • Worked in bugs & enhancement team, which involved resolving bugs and improving features according to client requirements.
  • Tech stack: Python, Flask, ElasticSearch, Redis

Perpule (Acqu. by Amazon)

May 2019 - July 2019 | Bangalore, India

  • I worked on the innovative cloud-based POS (point of sale) that enables offline stores in India to manage better their inventory, checkout process, and overall customer experience.
  • Developed APIs for real-time syncing of store and cloud, allowing clients to seamlessly checkout analytics impacting 30+ offline big multi-chain retail stores in India.
  • Tech stack: Python, Django, JavaScript, Angular6, Docker, Kubernetes

Google Summer of Code 2018

April 2018 - August 2018 | Remote

  • I worked on FOSSASIA's Badgeyay project, which enables event organizers to generate badges of events, impacting 100+ event organizers and 1000+ events.
  • Developed the Restful APIs in Python.
  • Implemented database models and schemas.
  • Developed the front end of the badge generator in EmberJS.
  • Tech stack: Python, JavaScript, EmberJS, Flask, PostgreSQL

Research Publications

Scalable declarative HEP analysis workflows for containerized compute clouds

Co-authored with Dr. Tibor Simko et al. DOI: 10.3389/fdata.2021.661501

  • The paper described a novel approach for experimental high-energy physics (HEP) data analyses centred around the declarative rather than imperative paradigm when describing computational tasks for analysis.
  • Contributed to the REANA platform development by improving support for several workflow specifications.

Ventures

Unus Imaginari

Co-Founder and Seed Investor - January 2022 - Present | India

Unus Imaginari is a modern, socially-conscious Indian streetwear brand inspired by quality craftsmanship, natural fibres & the power of human connection.

  • Co-founded a team of creative & imaginative minds who design soothing & aesthetic fashion clothes.
  • Managing the online shop and financial operations.
  • Brainstorming on brand values, Identity design system and consumer research.
  • Launched 50+ products with consistent 1.5 lacs (1.5k CHF) MRR (Updated: 2024).

Open Source Project Contributions

OpenEBS - Kubernetes storage simplified

April 2021 - June 2021 | Remote

  • OpenEBS is entirely native to Kubernetes and is implemented using microservices. OpenEBS can be installed via kubectl or Helm Chart and managed via Kubernetes custom resources.
  • Worked on improving the usability of OpenEBS and focused on implementing a user-friendly OpenEBS CLI (similar to kubectl) to perform various operations to get information from cStor volumes.
  • Tech stack: Golang

Rocket Software

May 2020 - July 2020 | Remote

  • I worked on Zowe desktop documentation viewer to implement a document viewing feature using a native VS code editor that impacts 4000+ Z-OS users.
  • Zowe provides modern interfaces created to host technologies that benefit the Z platform to modernize the Mainframe, simplify the user experience, and make the Mainframe more easily accessible.
  • Tech stack: JavaScript, React, ExpressJS

Software Consultancy

Codementor

August 2018 - September 2019 | Remote

  • Helped 100+ clients fix bugs and resolve Python-related issues through pair programming and live coding sessions.
  • Tech stack: Python, JavaScript

Conference Speaking

Web Accessibility In Django

PyCon Australia - 2019 | Sydney, Australia

Chirps: A Twitter Bot Framework in Python

EuroPython - 2019 | Basel, Switzerland

PyCon US - 2019 | Cleveland, US

Creating Badges with Badgeyay

FOSSASIA Summit - 2018 | Singapore

Education

MS in Computer Science: Semester Exchange Program

ETH Zurich - February 2025 - July 2025 | Zurich, Switzerland

  • Focused on cloud computing architecture and 3D vision.

  • Research project in the Computer Vision and Learning group on how LLMs and VLMs understand time-series data.

MS in Computer Science: Distributed Systems and Machine Intelligence

University of Basel - September 2023 - February 2026 | Basel, Switzerland

  • Focus areas: Distributed Systems and Machine Intelligence.

  • Courses: Pattern Recognition, Multimedia Retrieval, Privacy-Preserving Machine Learning

  • Master thesis, privacy-preserving ML: medical segmentation models leak patient data, but standard differential privacy destroys their accuracy. Trained UNet-family and RETFound foundation models under DP on retinal OCT and lung CT, and benchmarked advanced clipping and morphology mechanisms that recover much of the lost accuracy.

Entrepreneurship: From Research to Startup Success

Talent Kick : ETH Zurich - February 2024 - August 2024 | Zurich, Switzerland

  • Participated in a dynamic seminar to convert research innovations into viable startups.

  • The curriculum encompasses building interdisciplinary teams, engaging customers, crafting MVPs, market identification, and effective scaling strategies, with guidance from industry leaders.

B. Technology: Computer and Communication Engineering

The LNM Institute of Information Technology - 2016 - 2020 | Jaipur, India

  • Studied various subjects from Data structures, Algorithms, Probability, Statistics, Operating Systems, Software Engineering, Data Science & Artificial Intelligence, Information Visualisation, Computer Networks & Database management systems.

  • Accomplishments

    • CNCF Scholarship
    • Linux Foundation Scholarship
    • Facebook Udacity Scholar
    • Google Udacity Scholar
    • Google Code-In Mentor 201{7,8}
    • 1st Place - FOSSASIA Codeheat Coding Competition 2018
    • 1st Place - UNESCO Open Science and Data Hackathon 2018
    • 4th Place - LNM Hacks 2.0 Hackathon 2017

Non Tech Work

FreeCodeCamp & Hackernoon

Technical Writer - December 2017 - March 2018 | Jaipur, India

Excellion Capital

Video Editor - June 2017 - July 2017 | Jaipur, India

  • Created promotional videos for the company, which involved working with Adobe Premiere Pro.

Kydz Adda

Designer - January 2017 - March 2017 | Bangalore, India

  • I have created cool designs for Kids, which involved working with Adobe Illustrator and Photoshop.

Thanks for reading until here :)