Hrishikesh Jadhav
AI Engineer, Germany
Open to AI engineering roles in Germany from October 2026, with a German residence and work permit.
Experience
- 10/2025–09/2026
AI Engineer, Toku
Global employer-of-record and payroll infrastructure platform, remote-first.
- Replaced a manual payroll data upload with a self-hosted LLM extraction pipeline (Python, FastAPI, PostgreSQL, DigitalOcean) that masks personal data before inference, cutting hours of manual handling per payroll cycle down to minutes.Self-hosted so that no employee data leaves controlled infrastructure.
- Kept field-level extraction accuracy above the agreed threshold, measured by per-field exact match against a golden evaluation set, by owning the evaluation harness and a CI regression gate that blocks any deploy below it.So a change that lowers extraction accuracy cannot reach payroll data.
- Brought inference spend down to a fraction of equivalent third-party API usage, measured by monthly infrastructure billing, by running open-weight models on self-hosted GPU infrastructure instead of a vendor API.
- Put a RAG assistant (FastAPI, LangChain, pgvector), an MCP server exposing internal systems to LLM clients, and n8n automations for onboarding, payslip processing and support triage into daily use across product, engineering and sales, then owned onboarding, documentation and hands-on support for non-technical users.
- Moved security review of code changes to pull-request time across thousands of pull requests by integrating LLM-assisted security review into GitHub CI for Toku's financial platform.
- Built the customer-facing help centre and AI assistant (toku.com/help, Next.js) on Toku's Notion knowledge base, covering payroll, benefits, policy and country-guide articles, with general-help and visa-support content in separate retrieval corpora and fallback across model providers.Separate corpora so immigration and work-authorisation questions never resolve against general payroll content.
- Built status.toku.com, the public status page: monitored components probed from the Cloudflare edge on a short fixed interval, automatic incident open and resolve, a dependency rollup across DigitalOcean, Cloudflare and PropelAuth, and a JSON API and RSS feed. New incidents notify the team on Slack, and failures that need a code fix are picked up by a coding agent.Probing the edge and the origin separately localises a failure to the CDN or the application before anyone looks at it.
Personal data is masked before inference, and a golden-set gate decides what deploys. - 04/2022–09/2025
Data Scientist, GfK - An NIQ Company
Working Student, Data Science, 04/2022–08/2024
- Cut manual effort on product-taxonomy classification by around 30% against the prior manual process, measured on production catalogue throughput, with an LLM classification service on Amazon Bedrock combining prompt engineering and RegEx post-processing.
- Cut catalogue lookup time by around 40% against the previous keyword search by building vector-embedding semantic retrieval over internal catalogues.
- Improved simultaneous-viewer prediction accuracy by around 20% over the incumbent model, measured on held-out device-usage events, with a CatBoost model on 30+ engineered features from sociodemographics, temporal slots and programme metadata.
- Kept scoring pipelines running 24/7 without manual intervention by orchestrating S3 ingestion, feature generation, model scoring and automated integration tests in GitLab CI/CD, containerised in Docker on Linux.
- Built a data-fusion pipeline that clones TV-panel households using K-Nearest Neighbors, with a genetic algorithm selecting optimal subsets.
- 04/2020–08/2020
Software Developer (Internship), Sapio Analytics Pvt. Ltd.
- Built COVID-19 decision-support models in Python (SEIRD, scikit-learn, SciPy) and forecasting dashboards (Plotly, AWS) used to inform Government of India lockdown and testing-strategy decisions, with a 4.8% RMSE reduction against the prior baseline.
Projects
Internal RAG learning assistant
NIQ/GfK HACKFEST 2024, Top 3. A retrieval-augmented assistant over internal learning material with audio ingestion via Whisper, designed, built and demoed inside 24 hours.
Education
- 2022–2024
M.Sc. Artificial Intelligence Engineering (not completed)
- 04/2022–10/2022
Honours Degree in Entrepreneurship
- 2018–2021
B.E. Computer Science
Research
- 2025
- 2021
Awards
- 2024
NIQ/GfK HACKFEST · Top 3, with a RAG learning assistant built and demoed in 24 hours
- 2024
BMW Innovation Challenge · Selected participant, 24-hour challenge at BMW iFactory Dingolfing (DocCheck use case)
- 2020
IEEE Machine Learning Hackathon · 1st place
- 2020
HackCovid-19 · Winner among 130 teams
- 2017
Smart India Hackathon · Winner (Ministry of Defence)
Skills
- LLM and GenAI
- LLM application design, prompt engineering, RAG, vector search, evaluation and regression testing, guardrails, LangChain, LangGraph, CrewAI, n8n, Claude SDK, OpenAI SDK, OpenRouter, Amazon Bedrock, Whisper, Hugging Face
- Engineering
- Python, SQL, JavaScript, TypeScript, REST API design, FastAPI, Flask, Django, React, Next.js
- ML
- PyTorch, scikit-learn, CatBoost, pandas, NumPy, feature engineering, model evaluation
- Data and retrieval
- Vector embeddings, PostgreSQL, pgvector, Qdrant, FAISS, Pinecone, data-processing pipeline design, Amazon S3
- Cloud and DevOps
- AWS (Bedrock, S3), Google Cloud Platform, DigitalOcean, Docker, GitLab CI/CD, Cloudflare, Linux, Git
- Languages
- English (C2), German (A1 certificate, B1 course attended), Hindi (native), Marathi (native)
Print this page to save it as a PDF.