Machine Learning Engineer @ BlackRock

Tamogh

I'm a Machine Learning Engineer

Building unified ML platforms at BlackRock — from open-source research to multi-cloud production inference.

01. About me

Building the future
one model at a time


I'm a Machine Learning Engineer at BlackRock building a unified ML platform for registering, training, tuning, and deploying models across multi-cloud environments at production scale.

My stack spans Rust, KServe, KEDA, Azure, and PyTorch. I was also a Google Summer of Code 2024 contributor, having integrated MCMC and Variational Inference into Neuroptimus for neuronal parameter estimation

Completed a B.E. in Computer Science at Thapar Institute of Engineering & Technology

tamogh@ml-workstation ~

$ cargo run --release -- deploy --model gpt2-ft

Building Docker runtime from user deps... ✓

Pushing to Azure Container Registry... ✓

Deploying via KServe on AKS...

Health check: pod/gpt2-ft-0 Running ✓

KEDA autoscaler: minReplicas=1 maxReplicas=20

Inference endpoint live → /v1/models/gpt2-ft ✓

Metrics streaming → Prometheus + Grafana

$


1+
Years at BlackRock
GSoC
2024 Contributor

02. Skills

Technical Arsenal


From research to deployment — the full ML lifecycle.


ML / AI

PyTorch93%
TensorFlow / Keras85%
Scikit-learn90%
Transformers / HuggingFace88%
OpenCV82%

Systems & Backend

Python95%
Rust80%
Scala72%
C / C++75%
gRPC / REST85%

Cloud & MLOps

Microsoft Azure85%
Docker / Kubernetes83%
KServe / KEDA80%
Apache Spark / Airflow78%
PostgreSQL / MinIO80%

Arch. & Design

Domain-Driven Design85%
Microservices88%
SOLID Principles90%
Event-Driven Arch.80%
CQRS / Event Sourcing72%


Topics & Domains

TransformersGNNsBayesian MLMCMCVariational InferenceComputer VisionNLPETL PipelinesKServeKEDACI/CDHadoopLuigiGrafanaPrometheustusdMinIOS3DDDMicroservicesSOLIDClean ArchitectureHexagonal Arch.Event-DrivenCQRSEvent SourcingAPI-First Design12-Factor App

03. Projects

Things I've Built

Production ML systems, research prototypes, and everything in between.

2025

Unified ML Platform — BlackRock

Production ML platform to register, train, tune, evaluate, and deploy open-source, proprietary, and third-party models across multi-cloud. Backend in Rust with DDD, KServe for serving, KEDA for event-driven autoscaling. Delivers model inference-as-a-service with full lifecycle monitoring via Grafana & Prometheus.

Multi-cloudEvent-driven scaleFull lifecycle
RustKServeKEDAAzureDockerPostgreSQLMinIO
2025

ETL Pipeline Framework — BlackRock

Production-grade ETL pipelines for financial data forming the backbone of a trade library. Built a test server framework enabling rapid creation and validation of pipelines without manual environment setup. Integrated LLM-based capabilities into legacy data workflows.

Financial dataAuto test frameworkLLM-augmented
PythonSparkAirflowLuigiLLMPostgreSQL
2024

Bayesian Inference — GSoC 2024

Integrated MCMC and Variational Inference into Neuroptimus for neuronal parameter estimation under Dr. Sbalocz Kali at Google Summer of Code. Implemented global optimization techniques and custom loss functions for improved convergence.

Google GSoCBayesian MLNeuronal Params
PythonMCMCVariational InferenceNeuroptimusSciPy
2024

Chemical Reaction Prediction

Hybrid architecture combining Graph Attention Networks and Transformers to predict chemical reaction outcomes. Built local + global molecular embeddings capturing both structural and relational properties. Achieved 74% prediction accuracy.

74% accuracyHybrid GAT+TXMay–Nov 2024
PyTorchGATTransformersGNNsRDKit
2022

Soil Moisture Prediction — ThaparSat / ISRO

CNN-based soil moisture prediction model trained on satellite-derived features under the ISRO program. Includes denoising algorithms for satellite-retrieved data and payload compression pipelines.

ISRO ProgramCNN + SatelliteSince 2022
PyTorchCNNSatellite DataDenoisingCompression
2023

Motion Amplification & Frequency Analysis

Phase-based motion amplification and frequency extraction from video streams. Smart India Hackathon 2023 Finalist. Achieved 88% accuracy in micro-motion signal detection from raw video using FFT-based analysis.

88% accuracySIH 2023 FinalistReal-time
PythonOpenCVSignal ProcessingPhase-basedFFT

04. Experience

Where I've Worked

Machine Learning Engineer

BlackRock

Jul 2025 – Present

Building a unified ML platform to register, train, tune, evaluate, and deploy open-source, proprietary, and third-party models across multi-cloud environments. Architected backend systems using Domain-Driven Design in Rust. Developed a custom model deployment service generating Docker runtimes dynamically. Orchestrating production inference with KServe and KEDA. Monitoring via Grafana and Prometheus.

RustKServeKEDAAzureDDDMinIOPrometheus

Software Developer Intern

BlackRock

Jan 2025 – Jun 2025

Developed and optimized ETL pipelines for financial data, forming the backbone of a production-grade trade library. Modernized legacy systems by integrating LLM-based capabilities into data workflows. Built a test server framework for rapid creation and validation of ETL pipelines without manual setup.

PythonETLSparkLLMAirflowPostgreSQL

Open Source Developer (GSoC 2024)

Google Summer of Code · Neuroptimus

May 2024 – Aug 2024

Integrated Bayesian inference methods — MCMC and Variational Inference — into Neuroptimus for neuronal parameter estimation under mentor Dr. Sbalocz Kali. Implemented global optimization techniques and custom loss functions for improved convergence and accuracy.

PythonMCMCVariational InferenceBayesian MLNeuroptimus

Student Engineer Intern — ThaparSat

ISRO Program · Thapar University

May 2022 – Jul 2024

Developed a soil moisture prediction model based on CNNs and other architectures using satellite-derived features under mentor Dr. Mamta Gulati. Worked on denoising algorithms for satellite-retrieved data and payload compression pipelines.

PyTorchCNNSatelliteDenoisingCompression

05. Contact

Let's Collaborate