Berlin · Remote · Open to ML Engineering, Forward Deployed ML & Technical Leadership Roles

I build production
ML systems.

8+ years taking models from data and training through APIs, deployment, observability, and product experiments. Recommendation, personalization, search, speech, and computer vision.

Modelling · Systems · Product
View Experience LinkedIn →
ship. learn. iterate.
8+ years
Production machine learning
+2.5-5%
Conversion across offer-ranking experiments
+50%
CTR over static recommendations
-90%
Manual image selection effort
About

I work where modelling meets production software.

Hi, I'm Kamal. I build ML systems end to end: datasets, features, training, evaluation, APIs, deployment, monitoring, and iteration with product teams. My background spans recommender systems, ranking, uplift modelling, computer vision, speech, embeddings, and LLM applications.

I work directly with product managers, project leads, and domain stakeholders to turn ambiguous problems into testable ML solutions, production integrations, and measurable outcomes. I communicate technical tradeoffs, support adoption after launch, and mentor engineers and data scientists. Outside work, I ship my own mobile and web products. I have published three iOS apps and a paper at ECCV.

MSc CS · Univ. Freiburg · GPA 1.5 Berlin, Germany · Remote ECCV Publication 3 App Store apps
Experience
2021 to present
HolidayCheck GmbH
Senior Data Scientist · ML Engineering & Personalization

Lead end-to-end ML work across modelling, production systems, experimentation, and product adoption.

Personalization, ranking, and decision systems
  • Built user-intelligence services combining identity, behavioural history, booking intent, affinities, journey stage, and price sensitivity
  • Developed recommendation, ranking, propensity, and uplift models for offer selection and personalization
  • Delivered 2.5-5% conversion uplift across offer-ranking experiments
Model development and ML platform
  • Built training and evaluation pipelines for offer ranking, booking propensity, and uplift modelling
  • Standardized experiment tracking, model registry, and artifact workflows with MLflow on Kubernetes
  • Owned FastAPI inference services, Airflow orchestration, observability, and production support
Multimodal embeddings and search
  • Built transformer-based embedding and vector-search pipelines for images, reviews, and travel Q&A
  • Developed backfill and daily-ingestion workflows with asynchronous inference
  • Improved production reliability through memory tuning, resource management, and failure handling
Stakeholder delivery and technical leadership
  • Worked directly with product managers, project leads, and domain stakeholders from problem discovery through production rollout
  • Translated business requirements and operational constraints into ML roadmaps, system designs, APIs, and testable experiments
  • Communicated model tradeoffs, technical constraints, and experiment results to technical and non-technical audiences
  • Supported adoption after launch and incorporated stakeholder feedback into model and system iterations
2020 to 2021
Amazon
Applied Scientist
  • Shipped second-pass language-model rescoring for Alexa German ASR, reducing word error rate for millions of users
2018 to 2020
HolidayCheck GmbH
Data Scientist
  • Built a personalized travel recommender with 50% higher CTR than the static baseline
  • Productionized neural image assessment, cutting manual selection effort by 90%
  • Developed a multilingual conversational system using sentence embeddings
2015 to 2018
Univ. Freiburg & Uniklinikum
Research Assistant
  • ML for debate-video analysis using real-time audience response data
  • Built a virtual-microscopy learning tool for medical education
2014 to 2015
Infosys · Chandigarh, India
System Engineer
  • Developed and supported enterprise banking software before moving into machine learning
Published on the App Store
VoiceDoc transcription screen
VoiceDoc
Record, transcribe, and transform voice into structured documents entirely on device. Local AI, 100+ languages, no cloud.
App Store ↗
NimbleDay priority planning screen
NimbleDay
Energy-based task & goal planner with weekly goals, focus sessions, habits and iCloud sync.
App Store ↗
Sehaj Breath meditation screen
Sehaj Breath
Calm breathwork, guided meditation and emotion journaling. No streaks, no pressure.
App Store ↗
Research & Certifications
ECO: Efficient Convolutional Network for Online Video Understanding · ECCV arxiv →
Deep Reinforcement Learning Nanodegree
Computer Vision Nanodegree
Natural Language Processing Nanodegree
TensorFlow Developer
Machine Learning on Google Cloud
Deep Learning Specialization
In Development
01
LeadBloom

Evidence-first lead intelligence. AI agents discover buying signals across Reddit, Hacker News, GitHub, Product Hunt & 7+ more platforms. Full-stack SaaS with multi-tenant auth.

Next.js · FastAPI · Supabase · Gemini LLM
View live product ↗
02
High Intent Lead Engine

Agentic B2B sales engine. Profiles companies, ranks accounts by intent signals, and generates personalized outreach through an autonomous pipeline.

Next.js · Python Agents · LLMs · Supabase
How I Build
Model
Recommendation & ranking Classification & uplift Embeddings & vector search Computer vision & NLP LLM applications
Build
Python & SQL FastAPI services Training & evaluation pipelines Async inference API & system design
Operate
MLflow & Airflow Kubernetes & Docker AWS & GCP Tracing & Prometheus Production debugging
Deliver & Lead
Stakeholder discovery Requirements translation Solution design Product integration Experiment design Technical communication Mentoring

Let's
talk.

Open to Senior or Staff ML Engineer, Applied ML Engineer, ML Platform Engineer, Personalization Engineer, Forward Deployed ML Engineer, and ML Tech Lead roles in Europe.

[email protected] linkedin.com/in/singh4493