Tejasvi | Full-Stack Engineer

Tech Stack

Languages

JavaPythonTypeScriptJavaScriptC++SQLShellBash

Backend

Spring BootHibernateJPANode.jsNext.jsExpress.jsHono.jsFastAPI

Frontend

React.jsNext.jsAngular.jsReact NativeReduxTailwind CSSShadcn UI

Databases & Messaging

PostgreSQLMySQLMongoDBRedisKafkaZookeeperElasticsearchDrizzle ORMPrisma

Cloud & DevOps

AWS (ECS, S3, EC2)CloudWatchDockerKubernetesCI/CDJenkinsGitHub ActionsMicroservices

Tools & ML/AI

GitRESTgRPCWebSocketsJWTLinuxPyTorchTensorFlowRAGLangChainGenAI

Experience

Full Stack Engineer

Limelight Software · Toronto, ON

Oct 2025 -- PresentCurrent

[SYNC]Engineered a unified OLAP state synchronization layer bridging in-memory cubes with bulk reload APIs, enforcing a strict backend single-source-of-truth model that prevented cross-session data divergence

[STABILITY]Stabilized multi-dimensional query execution with pre-validation guards on OLAP INTERSECT/search endpoints, eliminating recurring Bad Request cascades without adding latency

[PIPELINE]Refactored backend orchestration into a deterministic Persist → Cache Mutate → Publish pipeline, decoupling transactional writes from async eventing

[CONCURRENCY]Eliminated race conditions, duplicate cache invalidations, and non-deterministic state during high-volume reloads, restoring thread-safe execution

[PARSING]Extended ANTLR-based parsing and AST evaluation with dependency-aware stale marking, deterministically re-triggering formula recomputation across multi-threaded pipelines

Full-Stack Developer Intern

Hilo Design · Remote

Nov 2022 -- May 2023Completed

[LATENCY]Optimized Node.js/MySQL backend to achieve sub-200ms (p-95) response times

[SCALING]Deployed Dockerized microservices on AWS ECS with CloudWatch monitoring

[CACHING]Implemented Redis caching strategies, reducing DB query latency by 30%

Projects

#01

Distributed Leader Election Service

Fault-tolerant leader election module using Java and Apache Zookeeper, leveraging ephemeral sequential znodes and Watcher events within a Spring Boot framework for robust distributed consensus.

JavaSpring BootZookeeperKafkaMaven

#02

Learning Management System

Full-stack LMS using Next.js 14 (App Router) & TypeScript, featuring SSR/ISR, Stripe payments, Mux video streaming, and PostgreSQL via Drizzle ORM. Drag-and-drop course UI secured by Clerk RBAC.

Next.jsTypeScriptPostgreSQLStripeMux

#03

Scalable Chat Application

Distributed, real-time chat backend using TypeScript, Socket.io, and Kafka for high-throughput fault-tolerant message queuing, storing messages in PostgreSQL. Dockerized for 10K+ msg/sec at <100ms P99.

TypeScriptReactSocket.ioKafkaPostgreSQLRedis

#04

VLM Number Plate Recognition

Fine-tuned a 330M TrOCR (ViT encoder/RoBERTa decoder) on the CENPARMI LPR dataset, achieving 86.8% accuracy (2.3% CER) with sub-1s inference via a custom SafeForwardModel wrapper and data collator.

PyTorchTransformers (ViT/RoBERTa)FlaskAttention

#05

Multimodal Vision Transformer (PaliGemma/Siglip)

Siglip ViT & PaliGemma multimodal model in PyTorch, covering patch/positional embeddings, multi-head self-attention, and an inference pipeline with KV cache and GPU-accelerated Docker deployment.

PyTorchViTGemmaDocker

#06

LLaMA 2 Implementation

LLaMA 2 architected from scratch in PyTorch, implementing RoPE, GQA, and RMS normalization, with top-p sampling, temperature scaling, and efficient KV caching for fast autoregressive inference.

PyTorchLLaMA StackDocker

Education

Masters in Computer Science (Applied)

Concordia University · Montreal, QC

Jan 2024 -- May 2025AI/ML, Scalable Systems, Full-Stack Engineering

B.Tech. Computer Science & Engineering

Kurukshetra University · Kurukshetra, India

Aug 2019 -- Sep 2023Software Engineering, Data Structures, Algorithms

Get in Touch

Based in Toronto, ON, Canada. Open to full-stack, backend, and ML/AI engineering roles — reach out at t3jasvii@gmail.com or use the form below.