ML Systems
Training infrastructure, transformer systems, autograd, distributed runtimes, inference workflows, and the constraints that shape model execution.
Distributed TrainingAutogradTransformersInference
About
Omprakash Sahani
ML Systems Engineer · Software Engineer · Distributed Systems
I’m a computer science graduate who enjoys understanding how machine-learning systems behave beyond the model itself—how memory, communication, latency, observability, and reproducibility shape the way systems perform in practice.
Since 2024, I’ve been building independent projects across autograd, transformer infrastructure, distributed training, benchmarking, search evaluation, reproducibility, and robotics-data analysis. I like working from first principles and turning those ideas into tested, documented tools that make complex systems easier to inspect.
I’m looking for software engineering and ML systems roles where I can keep learning, contribute to reliable infrastructure, and help researchers and engineers understand and improve the systems they depend on.
What I Work On
Software Engineering
Backend systems, APIs, developer tooling, testing, automation, maintainable architecture, and reliable workflows that turn technical ideas into usable software.
Backend SystemsDeveloper ToolingTestingSystem Design
Performance Engineering
Benchmarking, profiling, observability, regression detection, memory analysis, communication overhead, and reproducible evaluation.
BenchmarkingProfilingObservabilityReproducibility
Robotics Data Systems
State-space analysis, robot trajectories, workspace coverage, dataset diagnostics, and interactive tools for understanding embodied-AI data.
LeRobotForward KinematicsDataset Analysis3D Visualization
Experience
2024 — Present
Independent Engineering Work
Designing and building independent projects across ML systems, distributed training, software engineering, performance analysis, benchmarking, search evaluation, reproducibility, and robotics data.
Approach
First-principles implementation, testing, documentation, benchmarking, and public case studies.
Selected Work
Built Atlas AI, an ML systems platform spanning autograd, transformer infrastructure, distributed runtime concepts, inference serving, observability, and benchmarking.
Built LeRobot State Atlas, which transforms LeRobot trajectories through URDF-based forward kinematics into dual-arm workspace coverage and interactive three-dimensional diagnostics.
Built SearchEval Lab for evaluating TF-IDF, BM25, and hybrid retrieval systems with relevance metrics, latency analysis, regression detection, and query-level failure reporting.
Built tools for distributed-training profiling, benchmark-regression detection, reproducibility auditing, and automatic differentiation.
Maintain documented and tested public engineering projects with case studies, architecture views, validation evidence, and live demonstrations.
Education
Bachelor of Technology
Computer Science and Engineering
Sanjay Ghodawat University
2020–2023
CGPA: 8.40/10 · First Class with Distinction
Diploma
Computer Engineering
Maharashtra State Board of Technical Education
2017–2020
First Class