Skip to content
View Shibin506's full-sized avatar

Highlights

  • Pro

Block or report Shibin506

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. 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
Shibin506/README.md

Shibin Biji Thomas

Software Engineer · AI/ML · Full-Stack · Cloud Backend

Hands-on experience in AI/ML, full-stack and cloud backend engineering, from shipping production systems at LinkedIn and Accenture.

AI/ML & Software Engineer Intern @ LinkedIn  ·  ex-Software Engineer @ Accenture (3.5 yrs)  ·  M.S. Applied Data Intelligence @ San José State University (Dec 2026)  ·  San Francisco Bay Area, CA

LinkedIn Email Portfolio

Open to new-grad Software Engineering roles · MS graduating December 2026


About me

  • 🤖 At LinkedIn I worked on the feed recommendation system, shipping ranking-model improvements in PyTorch and TensorFlow (11 of 12 AUC wins), along with RAG and LLM evaluation pipelines.
  • 🧱 Before that I spent three and a half years at Accenture as a full-stack engineer, building React and TypeScript apps on Go, Java and Python microservices on AWS, and raising platform availability from 98.2% to 99.7%.
  • 🌐 I contribute to rocketride-server, a ⭐ 17.8k open-source C++/Python AI pipeline engine, fixing bugs in its CLI, LLM provider integrations, and database and graph nodes.

Experience

💼 AI/ML & Software Engineer Intern — LinkedIn  ·  Sunnyvale, CA  ·  May 2026 – Aug 2026

  • Delivered production ranking-model optimizations for the feed recommendation system in PyTorch and TensorFlow across concurrent pipelines: 11/12 AUC wins (up to +0.20% Click) and 11/12 Normalized Entropy (NE) improvements.
  • Built custom RAG pipelines, LLM evaluation workflows and AI agent frameworks in Python and PyTorch, using vector databases to raise prediction accuracy from 72% to 85%.
  • Engineered automated cloud telemetry, workflow automation and audit monitoring pipelines on AWS (SNS → Firehose → S3, Lambda) and GCP in Go and Python, cutting manual reporting effort by 90%.
  • Developed high-throughput serverless microservices and backend APIs in Go and Python, optimizing stream processing and event feedback loops to cut latency from 120 ms to 85 ms.
  • Used AI tools (Claude, Copilot) and serverless AWS Lambda routines for rapid service prototyping, raising execution reliability from 65% to 87%.
  • Ran containerized services on AWS EKS and Kubernetes with Docker, Airflow DAGs and Git, with automated unit tests and CI/CD verification gates.

💼 Software Engineer (Full-Stack) — Accenture  ·  Pune, India  ·  Jul 2021 – Dec 2024

  • Architected scalable serverless microservices and backend transaction systems in Go, Java and Python, lifting platform availability from 98.2% to 99.7%.
  • Built client web apps and integrations in React.js, TypeScript and REST APIs, and designed gRPC contracts and GraphQL schemas with Protobuf to standardize data flow.
  • Tuned PostgreSQL schemas, indexes, queries and connection pooling, cutting backend API response times from 450 ms to 280 ms.
  • Automated deployments with Docker, Kubernetes and Jenkins on AWS and GCP, with live telemetry monitoring and automated tests for zero-downtime releases.

Open source

Contributor to rocketride-org/rocketride-server (⭐ 17.8k), an open-source, multithreaded C++17/Python engine that runs AI workflows across 130+ Python nodes, 15+ LLM providers and databases.

PR What it fixes Status
#2679 A bare --modelserver port (e.g. --modelserver=5590) was read as a hostname; it now means localhost:PORT. Passed CI on macOS, Ubuntu and Windows ✅ Merged
#2680 The Gemini LLM node now actually sends the configured output-token limit to Google and saves it where it is read back; image models capped at Google's published limits 🔍 In review
#3276 With the graph execute tool disabled, agents now get pointed to the read tools (get_data, query) instead of a dead-end error. Covers FalkorDB, Neo4j and RocketRide Graph 🔍 In review
#3277 An unfilled ${ROCKETRIDE_*} credential in a database or graph node now stops the run and names the missing variable, across 7 database and graph nodes; 19 new tests 🔍 In review

Featured projects

A GitHub App that reads a failed CI run on a pull request and posts the root cause and a fix.

Mender demo: a failing test is diagnosed, the fix is verified, and the tests go green

  • Diagnoses failures with an LLM (Claude or Gemini, typed structured output) and shows a fix only if it applies to the PR's head commit; locally it also runs the tests on the patched code and hides fixes that still fail.
  • Broker-less Postgres job queue (SKIP LOCKED) with webhook signature checks, deduplication, retries with backoff and crashed-job recovery; a concurrency test shows exactly-once claiming across 8 workers.
  • Log extractor covering ~30 failure patterns across 8 toolchains sends ~0.3% of each log to the model; on a 30-case eval, Gemini 3.5 Flash got the failure type right in 25 of 26 answered cases and named the file to fix in 17 of 19.

Python FastAPI PostgreSQL Claude API Gemini API GitHub Apps SQLAlchemy Docker Prometheus GitHub Actions

Upload a meeting transcript, get decisions and action items linked to the exact timestamp, then ask questions across all meetings.

Watch the 1:45 narrated Minutes demo on YouTube

▶️ Watch the 1:45 narrated demo on YouTube

  • Summary, decisions and action items each link to the transcript timestamp; the model cites segment numbers, so timestamps cannot drift.
  • Streamed answers (server-sent events) with numbered citations, retrieved from an SQLite FTS5 BM25 index; retrieval scores hit@3 = 1.00 on a 20-question eval that runs in CI.

Python FastAPI SQLite FTS5 Claude API Gemini API React TypeScript Docker GitHub Actions

Client apps stream events and documents in; a lake pipeline makes them fast to query; an assistant answers from the documents with citations.

  • A Go gateway ingests events over gRPC and REST with per-tenant JSON Schema validation (~78K events/s from one client) into an S3 data lake; a Java GraphQL API versions schemas and rejects breaking changes.
  • A nightly Airflow pipeline runs Spark compaction to partitioned Parquet and Trino rollups with row-count reconciliation, cutting dashboard query latency by 70% over 2.35M events.
  • RAG assistant with hybrid pgvector + full-text retrieval, cited streaming answers and feedback-aware ranking (0.89 hit@3 on a 26-question eval), plus a React + TypeScript dashboard.

Go Java Python TypeScript React GraphQL gRPC Spark Trino Airflow PostgreSQL pgvector S3 Docker

A log-structured merge tree, the design behind LevelDB, RocksDB and Cassandra.

  • Checksummed write-ahead log, SSTables with Bloom filters and block index, levelled background compaction, and MVCC snapshots with range scans that stay consistent while writers and compaction run.
  • Crash tests kill a real process at every step of every disk operation.
  • On 1M keys: level with RocksDB on random writes (381K vs 386K ops/s) and 1.6x faster on in-memory point reads.

Rust LSM Tree Write-Ahead Log MVCC Bloom Filters Criterion


More projects

Project What it is Stack
redos-proof Finds regexes vulnerable to catastrophic backtracking (ReDoS) in JS / TS / Python / Ruby code and proves each finding by running the generated attack on the real engines; 100% recall, 0 false positives on a self-verifying corpus. CLI, pre-commit hook, GitHub Action and SARIF output Rust · NFA modelling · Static analysis · SARIF
MedBridge Turns hospital discharge papers into a plan patients can follow: an LLM extracts medicines, appointments and warning signs, and code verifies every item against the original. Plans in 8 languages, daily SMS / Telegram check-ins that escalate to a care-team dashboard; 804 tests, incl. a 336-case safety eval Python · FastAPI · Next.js · TypeScript · LLMs · Twilio · Telegram
ScreenSensei (team project, Cal Hacks 13.0) Desktop AI tutor that reads your screen, points and draws on it, and talks you through math, chess, software and languages step by step; narration starts ~3 s after a question while the rest of the lesson streams in. 380+ unit tests and a 13-case accuracy eval TypeScript · Electron · React · Fastify · Claude · Cartesia · Convex
Autonomous AI Agent & Serverless Workflow Engine Agent framework and serverless workflow engine with custom RAG pipelines and Kafka; event-driven streaming in Apache Flink cut real-time queuing latency from 150 ms to 95 ms, on AWS EKS at 99.5% node reliability Go · Python · Rust · Kafka · Flink · AWS Lambda · Kubernetes
BookTable 🏆 2nd Prize, SJ Hacks Restaurant search, reservations and real-time availability: Spring Boot microservices on MongoDB with OAuth2 / JWT / RBAC, a React frontend, and CI/CD to AWS (EC2, ALB, Auto Scaling) Java · Spring Boot · React · MongoDB · AWS
GlobalGuide AI Travel Planner Agentic travel planner on LangGraph that reasons across 9 custom tools (weather, places, currency) to build personalized itineraries, served by FastAPI Python · LangGraph · Groq LLM · FastAPI · Streamlit
AI Medical Chatbot (RAG) RAG chatbot that retrieves and summarizes medical documents in real time; 91% semantic retrieval accuracy over 5K+ healthcare records Python · HuggingFace · FAISS · LangChain · Mistral 7B
Spotify Data Warehouse End-to-end warehouse on Spotify API data; a Snowflake star schema made genre / artist queries 30% faster; 20+ visualizations Python · AWS S3 / Glue · Snowflake · Power BI
Real-time Yelp Sentiment Streaming ingest with Apache NiFi, PySpark sentiment classification of 5,000+ reviews at 90%+ accuracy, S3 history and a live Streamlit dashboard Python · NiFi · PySpark · VADER / TextBlob · AWS S3

Skills

Area Tools
Languages Go · Python · TypeScript · JavaScript · Java · C++ · Rust · SQL · Bash
AI / ML PyTorch · TensorFlow · Recommendation Systems · Feed Ranking · RAG Pipelines · LLMs · LLM Evaluation · LangChain · Vector Databases · AI Agents
Full-stack & APIs React · Next.js · Node.js · REST · gRPC · Protobuf · GraphQL · WebSockets · Microservices
Data & cloud PostgreSQL · DynamoDB · Redis · Cassandra · Vector Search · AWS · GCP
Streaming & processing Kafka · Apache Flink · Apache Spark · Trino · Airflow · Serverless · Distributed Systems
DevOps & testing Docker · Kubernetes · Git · CI/CD · Jenkins · GitHub Actions · PyTest · Linux · System Monitoring

Education

🎓 M.S. Applied Data Intelligence, San José State University · Jan 2025 – Dec 2026 (expected) · GPA 3.5
Coursework: Distributed Systems · Database Management · Algorithms · Cloud Infrastructure · Machine Learning


Contact

📫 shibinbiji2@gmail.com  ·  LinkedIn  ·  Portfolio  ·  San Francisco Bay Area, CA

Open to new-grad Software Engineering roles (AI/ML, full-stack, backend) · MS graduating December 2026.

Pinned Loading

  1. Spotifyzer Spotifyzer Public

    This is the repository for data warehouse project for Spotifyzer analysis using AWS S3 , AWS glue, Snowflake Python, Pandas, Numpy. It integrates ETL pipelines, dimensional modeling, and analytics …

    Jupyter Notebook

  2. AI-MEDICAL-CHATBOT-with-RAG AI-MEDICAL-CHATBOT-with-RAG Public

    Medical Q&A chatbot using RAG: HuggingFace embeddings + FAISS vector search over 5K+ healthcare records, answered by Mistral 7B via LangChain, with a Streamlit UI.

    Python

  3. GlobalGuide-AI-Travel-Planner GlobalGuide-AI-Travel-Planner Public

    Python 1

  4. Sentiment-Analysis-on-Yelp-Reviews Sentiment-Analysis-on-Yelp-Reviews Public

    Sentiment Analysis on Yelp Reviews

    Python