
Shubhang Sinha
Machine Learning Engineer · RL, Real2Sim2Real, Robot Policies, Post-training
About
Hi folks, I predominantly like working and reading things in the space of Post-training, Distributed Training and Real2Sim2Real pipelines. I currently work in PixelsTwo as an ML Research Engineer working on Research problems related to Real2Sim.
Experience
ML Research Intern · PixelsTwo
Oct 2026 — present- Built an end-to-end Real2Sim pipeline converting a single monocular/stereo video into a ready simulation scene.
- Building robust sim environments for better policy training.
- Building Talent Angels, a suite of AI graph agents that lets users explore skills, tasks, and occupations across global taxonomies (ESCO, O*NET, SFIA, BLS, Lightcast) through a natural-language interface.
- Building and traversing knowledge graphs (Neo4j, LangGraph) to map relationships between skills, tasks, and occupations across taxonomies.
Google Summer of Code · DeepChem↗
May 2026 — Aug 2026- Integrated the DNABERT tokenizer and DNABERT-2 model into DeepChem’s infrastructure, enabling genomic-sequence research from within DeepChem.
- Submitted multiple pull requests adding featurizers, unit tests, and docs while building benchmarking pipelines for evaluating the DNA foundation model.
ML Research Intern · Indilingo
Dec 2025 — Jan 2026- Improved the efficiency of translation models for rare Indian languages using the ReAct framework and multi-shot prompting.
Elsewhere
Read my posts, grab my resume, or just email me.