Skip to content
View ankitdhandharia's full-sized avatar
🎯
Focusing
🎯
Focusing
4 followers · 7 following

Achievements

Achievement: YOLOAchievement: Pair Extraordinairex2Achievement: Pull Sharkx2

Achievements

Achievement: YOLOAchievement: Pair Extraordinairex2Achievement: Pull Sharkx2

Organizations

@hiringfleet

Block or report ankitdhandharia

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ankitdhandharia/README.md

Typing SVG

Portfolio  LinkedIn  Email  AWS Certified  Profile Views


About Me

I am a software engineer focused on distributed systems, data platforms, and machine learning infrastructure. I enjoy tackling problems where correctness and runtime performance matter most: eliminating multithreaded race conditions, resolving execution graph edge cases, and building high-throughput data pipelines that remain reliable under heavy loads.

  • 🎓 M.S. Computer Science @ Stevens Institute of Technology (Hoboken, NJ — GPA 3.71 / 4.0 · Digital Degree)
  • ☁️ AWS Certified Data Engineer – Associate (Verify on Credly)
  • ⚙️ Systems & Distributed Computing: Concurrency safety, race-condition elimination, execution graph optimization, and low-latency communication runtimes.
  • 🌊 Data Engineering & Lakehouses: Production-grade Medallion architecture (Snowflake + dbt Core + S3), incremental loading, SCD Type 2 snapshots, and streaming pipelines (Kafka + Spark).
  • 🔬 Upstream Open-Source Engineering: Actively resolving runtime edge cases and bugs across foundational repositories including Google Protocol Buffers, Microsoft DeepSpeed, ONNX, and PyTorch Vision.
  • 🌐 Portfolio & Architecture Write-ups → ankitdhandharia.com

⚡ Upstream Open-Source Contributions

Real-time snapshot of public pull requests and upstream engineering contributions across external ecosystems, auto-updated via GitHub Actions:

Upstream Open Source Ecosystem

🧠 Machine Learning Runtimes & Distributed Systems

  • ONNX Reference Runtime ★ 21.6k · PR #8599

    Handled legacy Resize inputs in reference implementation and preserved exact nearest-neighbor sub-pixel coordinate roundings across opset 10/11 transitions.

  • Microsoft DeepSpeed ★ 43.2k · PR #8631

    Fixed AutoSP test suite expecting LLM decoder attention layers to be auto-wrapped with DistributedAttention, aligning tests with the intentional sequence parallelism operator architecture.

  • PyTorch Vision ★ 18.0k · PR #9677

    Guarded reference evaluate() sample-count checks against uninitialized default process groups when benchmarks run in standalone non-distributed mode.

  • DMLC XGBoost ★ 28.9k · PR #12613

    Added strict Python type annotations to config and tree regularization test suites and enabled mypy CI verification.

📊 Time-Series, Bayesian Optimization & Statistical Engines

  • Unit8 Darts ★ 9.6k · Merged #3217 Merged #3210

    Merged: Resolved multithreaded race condition where ModelMeta.__call__ stored creation arguments in class-level attributes, preventing collisions during parallel Optuna sweeps (n_jobs>1). Also documented window parameter in CauchyNLLScorer.

  • Meta BoTorch ★ 3.6k · PR #3411

    Introduced pluggable cat_kernel_factory in MixedSingleTaskGP to support domain-specific categorical kernels alongside continuous dimensions.

  • AWS GluonTS ★ 5.2k · PR #3351

    Prevented unintended CUDA context initialization by enforcing CPU predictor placement when training with accelerator="cpu".

  • CausalEstimate · Merged #148 & tsauditor · Merged #74

    Merged: Fixed unaligned DataFrame index in Matching._compute_effect that caused silent NaN calculations; added single-row anomaly edge-case test suites.

🔌 Core Protocols & Cloud Tooling

  • Google Protocol Buffers ★ 72.1k · PR #30114

    Included duplicate collision key in C# MapField.Add exception message to accelerate debugging in downstream cloud logging services.

  • AWS SageMaker Python SDK ★ 2.3k · PR #6430

    Fixed Local Container mode failure when an S3 input channel points to an individual file object instead of a prefix directory.

  • MonitorControl ★ 34.5k · PR #1913

    Documented troubleshooting resolutions for macOS dim screen capture conflicts and slider brightness jumps.

View All PRs


🛠 Featured Systems & Architecture

🏔 Medallion Data Lakehouse

dbt Core Snowflake AWS S3 Jinja

End-to-end cloud data warehouse implementing the Bronze → Silver → Gold architecture on an Airbnb dataset.

  • Incremental loading + SCD Type 2 snapshots for point-in-time dimensional accuracy
  • Metadata-driven Gold layer with custom dbt Jinja macros
  • Automated schema validation & freshness assertions — 40% compute cost cut via incremental materialization

⚡ Autonomous Multi-Source Aggregator

TypeScript Node.js Distributed Ingestion APIs

Distributed event ingestion and normalization platform integrating 8+ external sources (Luma, Meetup, Ticketmaster, Pretalx, Devpost, Kaggle).

  • Unified schema normalization and cross-platform deduplication
  • Zero-auth data adapters for open-data event syndication
  • Click-to-recommendation feedback loop with low-latency dispatch

🛡 Deterministic LLM Model Gateway

FastAPI Python OpenRouter Security Guardrails

Multi-tier LLM gateway with deterministic model dispatch (GPT-4o-mini & DeepSeek-V3.2) and token budgeting.

  • Shared daily cost/token caps across concurrent execution engines
  • Defense-in-depth security guardrails mitigating SSRF, prompt injection, and DoS attacks
  • Section-scoped semantic prompting for deterministic outputs

🤖 Progressive Hybrid RAG Pipeline

LangChain ChromaDB HuggingFace Vector Embeddings

Modular Retrieval-Augmented Generation system with dense embeddings and persistent vector indexing.

  • Zero cloud API dependency using local HuggingFace embeddings (384-dim)
  • Persistent ChromaDB vector store with sub-second retrieval across 400+ page documents
  • Clean OOP architecture ready for high-throughput inference

→ Browse all repositories


🛠 Technical Toolkit

Languages & Systems

Data Engineering & Big Data

Cloud & DevOps

AI / ML & Frameworks

Databases & Backend


🎓 Education & Credentials


Stevens Institute of Technology

Stevens Institute of Technology

M.S. in Computer Science
Hoboken, NJ · Sep 2024 – May 2026

GPA: 3.71 / 4.0

Distributed Systems · Machine Learning · Cloud Computing · Advanced Algorithms

🎓 View Digital Degree (Parchment) »


AWS Certified Data Engineer – Associate

Amazon Web Services

AWS Certified Data Engineer – Associate
Issued by Amazon Web Services Training & Certification

Active Credential: March 2026

🔗 Verify on Credly »


📊 GitHub Activity

Contribution Streak

Activity Graph


ankitdhandharia.com

Pinned Loading

  1. AWS_DBT_Snowflake AWS_DBT_Snowflake Public

    End-to-end data engineering pipeline using dbt Core, Snowflake & AWS S3 — implements the Medallion Architecture (Bronze → Silver → Gold) with incremental loading, SCD Type 2 snapshots, custom Jinja…

    Python

  2. Mini-Search-Engine Mini-Search-Engine Public

    This project implements a simplified search engine based on Section 23.6 of the textbook, designed to index and search HTML documents from a small website. The implementation uses inverted indexing…

    HTML

  3. Polars-vs-Pandas Polars-vs-Pandas Public

    Hands-on benchmark comparing Polars vs pandas for data-processing performance — speed, memory, and API ergonomics across common DataFrame operations.

    Jupyter Notebook

  4. RAG_Project RAG_Project Public

    A progressive RAG (Retrieval-Augmented Generation) project using LangChain, ChromaDB, HuggingFace Embeddings & OpenRouter — from plain text to PDF-based QA with persistent vector stores.

    Jupyter Notebook