Skip to content
Sharan Shyamsundar
Portrait of Sharan Shyamsundar

Founding AI Engineer · Germany

Hi, I'm Sharan. I build AI agents that do real work.

At qado, I build agents that help enterprise buyers keep the savings they negotiate. Before that, I built agents that do the books at Quid.

See my workBook 20 minutesDownload CV (PDF)

Selected work

Two build logs: what I built, what made it hard, what I learned.

qado · Procurement2026 – now

Agents that stop procurement savings from leaking away

Agents that read procurement contracts and catch where the savings slip away.

  1. contracts + invoices
  2. read the terms
  3. check invoices
  4. flag for the buyer
Read the build log →
Quid · Finance2025 – 2026

From LLM demo to agents that do the books

Taking early LLM prototypes to production agents for finance teams.

  1. receipt in Slack
  2. extract fields
  3. map categories
  4. expense report
Read the build log →

Background

Journey

  1. 2026 – nowFounding AI Engineer, qadoFounding team. Agents for enterprise procurement: contract compliance, overspend detection, negotiation prep.
  2. 2025 – 2026AI Engineer, QuidJoined early. Took LLM prototypes into production: multi-agent systems for accounting.
  3. 2023 – 2025Working Student, BASFSupply Chain Digitalisation team. Built Eddy, a RAG helpdesk for internal questions; supply chain dashboards.
  4. 2022 – 2024MSc Data Science, University of MannheimStudyMoved to Germany. Thesis on parameter-efficient fine-tuning of quantised LLMs.
  5. 2020 – 2023Quant Analyst & Consultant, Equity Data ScienceJoined the quant team early. ESG analytics products, automated investment workflows.
  6. 2020Management Trainee, Machine Learning, IT Mines TechnologyMy first machine-learning role in industry.
  7. 2019Data Scientist Intern, Saatchi & Saatchi · Publicis MediaInternship during my bachelor's.
  8. 2017 – 2020BSc Applied Statistics & Analytics, NMIMSStudyWhere numbers became a habit.

Recognition

  • Aug 2026Winner, AWS Challenge, JiVS Hackathon
  • Aug 2025Topic winner, JiVS Hackathon
  • Jan 20252nd overall + topic winner, JiVS AI Hackathon at WEF Davos
  • Aug 2024Topic winner, JiVS Hackathon
  • 2023Team winner, STADS Datathon, Mannheim

All team wins. JiVS hackathons with team Strike, organised by Data Migration International; the summer editions at Seeburg Castle.

How I build

  1. 01

    Learn the domain on purpose.

    I list what I know and what I don't, ask the people who've worked in it for years, and turn the gaps into a learning plan. AI helps: research, study plans, even a podcast of the paper I don't have time to read.

  2. 02

    Evals before features.

    If I can't measure it on real cases, it doesn't ship.

    See it in practice →
  3. 03

    The model is the easy part.

    Most of the work is the tools, context and checks around it.

  4. 04

    Agents earn their autonomy.

    Suggest, then approve, then act. Trust is measured before money moves.

  5. 05

    Try everything, keep what works.

    New models and coding agents land every few weeks. I test them early and share what helps with the team.

Other builds