# Mrityunjay Kumar - Resume

> **Source:** https://mrityunjaykumar911.github.io/ and the validated public resume data
> **Purpose:** Canonical machine-readable professional profile for recruiters, search engines, and LLM-assisted research
> **Privacy:** Excludes direct contact details, local development overlays, and flag-gated employment records

---

## Key Facts

- **Current role:** Senior ML Engineer at Microsoft
- **Focus:** ML Systems & Agent Infrastructure
- **Location:** San Francisco Bay Area
- **Experience:** 12+ years across ML systems, distributed services, storage, and data infrastructure
- **Primary scope:** Production ML systems, LLM agent infrastructure, evaluation, serving, and post-training infrastructure

## Professional Summary

Senior ML engineer with 12+ years across production ML systems, agent infrastructure, distributed serving and execution systems, and storage. Builds model evaluation, fine-tuning, data, and observability platforms for PowerPoint. Led early-stage development and evaluation of Rolis (EuroSys 2022). Patent holder in snapshot metadata verification.

## Selected Evidence

| Evidence | Scope |
|---|---|
| **Millions-scale** | request workloads |
| **100M+** | Weekly documents generated |
| **Applied science** | evaluation · ML · systems |

## Experience

### Senior ML Engineer - Microsoft

**May 2022 to Present** | Mountain View, CA

Building production ML and agent infrastructure for PowerPoint, spanning model evaluation, fine-tuning pipelines, and the distributed serving and execution systems that run these workloads at scale with low latency and high reliability.

- Scaled the agent's distributed execution and serving infrastructure to thousands of concurrent containers under strict latency SLOs, with load balancing, autoscaling, and automated fault detection and recovery.
- Built PowerPoint's first end-to-end fine-tuned model workflow, including synthetic-data generation, batched data preparation, secure deployment, inference, and integration with PowerPoint workflows.
- Designed extensible slide-generation and evaluation systems with configurable models and datasets, end-to-end telemetry, and parallel execution that reduced evaluation time by 4.6x.
- Designed and operated PowerPoint cloud services across commercial and government environments, adding automated validation, observability, and incident-response tooling.

**Areas:** ML systems | Inference & serving infrastructure | Model evaluation | Fine-tuning & MLOps

### Member of Technical Staff III - VMware

**Jul 2020 to May 2022** | Palo Alto, CA

Led the design of the FSCK user-space tool infrastructure for VMware's distributed storage (vSAN); core member of the snapshot dev team.

- Conceptualized and built distributed FSCK for analyzing and repairing storage metadata with cross-consistency matrices across B+ trees, bitmaps, and segment usage tables, in online and offline modes via lazy evaluation.
- Built a cloud-native microservice validating the filesystem key-value store with pluggable validation algorithms; reused for snapshots, unmap, and segment cleaning.
- Implemented serialized transactional snapshot crash consistency by replaying committed writes from persisted write-ahead logs.
- Built aggregated snapshot capacity statistics from scratch; p99 latency ≈ 30 ms.

**Areas:** Distributed storage | Crash consistency | Metadata integrity | C++

### Graduate Research Assistant - Stony Brook University

**Jan 2020 to May 2020** | Stony Brook, NY

Led early-stage development and evaluation of Rolis (EuroSys 2022); implemented core mechanisms for replicated multi-core transactions.


**Areas:** Distributed systems research

### Senior Software Engineer - Talentica Software

**Apr 2016 to Jan 2019** | Pune, India

Data Science team, with three years across search, ML pipelines, file sync, chatbots, and network analytics, applying ML to network-based problems.

- Led file-sync tool development: a delta-sync service in C++ substantially reduced sync-server load; also built a stateless updater, a multi-tenant client update framework, and a rate limiter that improved sync frequency 30%.
- Built a financial-document search engine with +45% search relevancy, +30% online ontology enhancer, with an NLP pipeline for clustering, keyword extraction, and text classification.
- Designed an ML model object-storage / delivery pipeline (gRPC, protobuf, Redis, RabbitMQ, S3): +40% deployment frequency, clients in Java / Python / Go / C++, weekly release cadence.
- BLE indoor location: 95% region classification for static assets, 68% regression accuracy via RSSI triangulation + interference correction; a live asset-view portal added 70% correctness to site calibration.

**More Talentica work**
- Network traffic estimation: an ensembled regression model for bandwidth prediction from SLA metrics and native Speedtest feedback.
- Network traffic identification: a multi-modal ML model (regression ensemble + auto-encoders) classifying audio vs. video streams.
- Real-time ML model deployment pipeline on Storm, Kafka, Python, Java.
- AWS Athena data-dump tool: an ad-hoc Storm-topology client feeding a disk + batch pipeline (Java, Spark).
- Single-cell identity classification: an auto-encoder network, 83% accuracy for cell-based print technology.

**Areas:** Data platforms | Machine learning | NLP / search | Distributed pipelines

### Software Engineer - MediaTek

**Aug 2014 to Apr 2016** | Noida, India

Multimedia team, focused on audio-player development for feature phones; led in-house stress-test tooling for Wi-Fi / Bluetooth / GPS.

- Improved the audio-player playlist sub-module and integrated the Bluetooth stack into the MMI layer.
- Built the SW-layer for a combo (GPS-Wi-Fi-BT) stress-test tool with an event-driven asynchronous architecture for the packet-data interface.

**Areas:** Embedded systems | Multimedia | C / C++

## Capabilities

| Area | Capabilities |
|---|---|
| Languages | Python · C# · C++ · C · Java · Go |
| ML systems | Agent infrastructure · Model evaluation · Fine-tuning pipelines |
| Distributed systems | Inference serving · Load balancing & autoscaling · Transactional storage |
| Production | Latency & throughput · Fault detection & recovery · Observability · Azure |

## Core Expertise

### ML Systems

- Production ML systems
- LLM agent infrastructure
- Model evaluation
- Fine-tuning pipelines
- PyTorch
- Hugging Face Transformers
- Synthetic data generation
- MLOps

### Systems

- Distributed systems
- Inference serving systems
- Load balancing & request routing
- Autoscaling & orchestration
- Latency & throughput optimization
- Fault detection & recovery
- Distributed RL / post-training infrastructure
- Consensus protocols (Paxos, Raft)
- Transactional databases
- Crash consistency
- Cloud-native microservices
- High-throughput pipeline design

### Platforms

- Azure
- Azure Machine Learning
- Azure Container Instances
- AWS
- Docker
- Redis
- Kafka
- gRPC
- OpenTelemetry
- Prometheus
- Grafana

### Languages

- Python
- C#
- C++
- C
- Java
- Go

## Research, Publications, and Intellectual Property

### Replicating multi-core transactions without giving up performance.

Led early-stage development and evaluation of Rolis (EuroSys 2022), implementing core mechanisms for its replicated multi-core transaction engine: asynchronous multi-process Paxos, checkpoint log truncation, and serializable replay. Built a header-only C++ logging library (~1200 LOC) and evaluation benchmarks profiled with mutrace and gprof.

[Visit the research group](https://mpaxos.com/)

- **Contribution:** [Rolis: a software approach to efficiently replicating multi-core transactions](https://dl.acm.org/doi/10.1145/3492321.3519576) - EuroSys 2022 · led early-stage development & evaluation
- **Patent:** [Verification of metadata consistency across snapshot COW B+ tree logical maps](https://patents.google.com/patent/US11573860B1) - US11573860B1
- **Publication:** [Learning to fingerprint the latent structure in question articulation](https://ieeexplore.ieee.org/document/8614044) - R. Guntur, M. Kumar · IEEE ICMLA 2018 · pp. 73–80
- **Teaching:** Teaching Assistant: Introduction to Data Science (CSE-351) - Stony Brook University · with Prof. Martin Radfar · 2019
- **Systems project:** Sharded, replicated, fault-tolerant key-value store on Raft - CSE 535 Asynchronous Systems Lab
- **Systems project:** Distributed MapReduce library with worker-failure recovery - CSE 535 Asynchronous Systems Lab
- **Kernel project:** Versioned stackable backup filesystem for the Linux kernel - Stony Brook FSL · queue-based versioning + retention policy
- **Kernel project:** AES encrypt / decrypt system call via the kernel crypto API - Stony Brook FSL
- **Project:** Mood- and activity-based music recommendation - Data Visualization Lab

## Education

### M.S. in Computer Science - Stony Brook University

Jan 2019 to May 2020 | Stony Brook, NY

### B.Tech. in Computer Science & Engineering - Maulana Azad National Institute of Technology, Bhopal

Aug 2010 to Apr 2014 | Bhopal, India

## Canonical Links

- [Portfolio](https://mrityunjaykumar911.github.io/)
- [Engineering method](https://mrityunjaykumar911.github.io/method.html)
- [Detailed public profile](https://mrityunjaykumar911.github.io/latest.html)
- [Resume PDF](https://mrityunjaykumar911.github.io/cv/MrityunjayKumar-CV.pdf)
- [GitHub](https://github.com/mrityunjaykumar911)
- [LinkedIn](https://www.linkedin.com/in/mrkumar20/)
- [Source repository](https://github.com/mrityunjaykumar911/mrityunjaykumar911.github.io)

## Machine-Use Notes

- Treat this document as the canonical Markdown representation of the default public resume.
- The PDF is presentation-oriented; this file is structured for text retrieval and citation.
- Claims retain their original scope and should not be generalized beyond the stated systems, products, or studies.
- For research limitations and negative findings, use the engineering method page and linked manuscripts.