A quick introduction — where I come from, what I've built, and why I'm equally at home in a distributed systems codebase, a GPU runtime kernel, a data warehouse, or technical architecture reviews.
I'm a software and systems engineer who builds across distributed systems, machine learning infrastructure, and high-performance runtimes. Currently finishing my Master of Science in Computer Science at Stevens Institute of Technology in Hoboken, NJ (graduating May 2026, GPA 3.71 / 4.0).
Much of my recent focus is dedicated to upstream open-source engineering. I contribute directly to foundational codebases with over 235K+ combined stars — including Google Protocol Buffers (C++ arena memory allocator fixes), Microsoft DeepSpeed (ZeRO-3 FP8 mixed-precision kernels), ONNX Runtime (dynamic shape inference validation for transformers), PyTorch Vision (NVDEC GPU video decoding deadlock resolution), and Unit8 Darts (probabilistic time series forecasting and model serialization).
My foundational training began with a BSc (Honours) in Information Technology specializing in Data Science from Somaiya Vidyavihar University, graduating with a 9.09/10 CGPA. That curriculum provided rigorous grounding in mathematical statistics, applied machine learning, algorithms, and database architecture.
At Stevens, I've expanded deep into distributed systems, advanced algorithms, and generative AI systems. Along the way, I earned the AWS Certified Data Engineer — Associate credential and three Azure certifications, formalizing expertise across cloud storage, streaming pipelines (Kafka, Spark), and data warehouse modeling.
Whether it's debugging a race condition in multi-GPU gradient binning or architecting production streaming pipelines with sub-second latencies, I care about correctness, determinism, and engineering rigor.
I'm open to full-time engineering roles and summer internships across Software Engineering, AI/ML Systems Engineering, Data Engineering, and Applied Data Science.
Distributed systems, runtime engines, and high-concurrency backends. Fluent in C++, Python, TypeScript, and modern engineering workflows. Experience across arena allocators, gRPC, and multi-threaded systems.
Focus on high-performance runtimes and training systems — ZeRO-3 sharding, FP8 execution, GPU kernel synchronization, ONNX dynamic shape validation, and RAG architectures.
End-to-end data pipelines, Medallion architectures, dbt transformations, Airflow orchestration, and Snowflake/cloud warehouses. AWS Certified Data Engineer Associate.
BSc Honours in Data Science (9.09/10) — rigorous training in mathematical statistics, hypothesis testing, double machine learning, causal inference, and probabilistic forecasting.
SQL-native metrics and KPI definition. Real-time dashboards, BI reporting, and automated analytics workflows that drive high-confidence decisions.
Statistical demand forecasting, operational workflow optimization, and translating complex distributed architectures into clear stakeholder deliverables.
Predictable, tested, well-documented. Production code nobody has to think about at 2am is worth ten "clever" solutions that demand constant attention.
Readable code, good PR etiquette, sensible git hygiene, tests that actually run. The multiplier effect of a clean codebase compounds across a team.
The other 90% is data pipelines, feature engineering, reproducibility, deployment, and monitoring. A 95% accurate model you can't deploy is worth less than an 85% model you can.
Schema validation, null checks, anomaly detection — built in from day one, not retrofitted after the first incident. Same goes for app validation and API contracts.
The best analyst I know says "a chart you have to explain is a chart that failed." I try to make every dashboard, metric, and model output communicate clearly without a 10-minute preamble.
Columnar storage, smart partitioning, efficient algorithms, right-sized infra — the best performance gains almost always double as cost savings. Same principle across app, ML, and data work.
Currently into Designing Data-Intensive Applications by Martin Kleppmann, The Pragmatic Programmer, and papers on distributed GPU training systems.
FP8 kernel optimization, vector DBs (ChromaDB, OpenSearch), multi-agent LLM systems, and the uv Python package manager.
Where runtime engineering meets ML acceleration — zero-copy memory patterns, custom allocators, and deterministic distributed training.
Full-time roles post-graduation (May 2026) and summer 2026 internships in SWE, AI/ML Systems, Data Eng, or Applied Data Science — NYC metro preferred but open to relocation.