text-3xl text-zinc-950text-zinc-50 font-medium
Nitin Prajwal
Creating with code. Small details matter.
Overview
Social Links
About
Hello, World.
I'm Nitin, a developer who thrives on building things that scale from cloud-native microservices to AI-driven systems. My work spans across Oracle projects, open-source contributions, and developer communities like GDG, where I've led teams and shipped real solutions.
I enjoy working where AI, cloud, and distributed systems meet, turning complex ideas into efficient, production-ready code. Whether it's fine-tuning LLMs, designing high-performance pipelines, or reverse-engineering browsers, I focus on engineering with clarity, speed, and impact.
Always exploring, always building.
Stack
Go
.NET
Firebase
Amazon Web Services
Kubernetes
Docker
Angular
Android Studio
Anaconda
Android
Arduino
Microsoft Azure
Canva
Google Chrome
C
CMake
C++
C#
Electron
Figma
Flutter
Git
Gradle
IntelliJ IDEA
JavaScript
Jupyter
Kotlin
Linux
MATLAB
MongoDB
MySQL
Next.js
Node.js
npm
Nuxt.js
OpenGL
PostgreSQL
PyTorch
Python
R
Raspberry Pi
React
Svelte
TensorFlow
TypeScript
Visual Studio Code
PuTTY
OpenCV
Jira
Java
HTML5
GraphQL
Fedora
Debian
Blender
Apache
Apache Kafka
Elasticsearch
Spring Boot
Prometheus
Grafana
Blog
Experience
HARMAN International
Current Employer- Built a production DevOps AI Platform in React/Next.js 16 and TypeScript, wiring OpenClaw with LangGraph to run autonomous multi-agent workflows; set up an OpenTelemetry pipeline that shipped LLM traces, metrics, and logs into Grafana/Prometheus/Loki for live observability over Kubernetes-hosted microservices deployed via ArgoCD.
- Wrote unit, regression, and E2E test suites in pytest and Jest, keeping code coverage above 90%; did code reviews, tracked defects end-to-end in JIRA, and shipped features in two-week Agile sprints with clear documentation.
- Wrote Go microservices with Kafka for async event processing and a GraphQL API layer; used Elasticsearch for log search and MCP (Model Context Protocol) to connect agents to internal tools — containerized everything with Docker and pushed to AWS ECR/EC2, cutting deployment time by roughly 45%.
- React
- Next.js
- TypeScript
- LangGraph
- OpenTelemetry
- Grafana
- Prometheus
- Loki
- Kubernetes
- ArgoCD
- Go
- Kafka
- GraphQL
- Elasticsearch
- Docker
- AWS
- pytest
- Jest
- JIRA
- Agile
Serent.ai
- Fine-tuned ModernBERT for Named Entity Recognition (NER) on unstructured clinical notes, achieving 85% precision on ICD-10 code prediction; built context and negation detection on top to filter "ruled-out" diagnoses, cutting manual chart review time by 70%.
- Fine-tuned a Qwen large language model to generate clinical reasoning chains and classify diagnoses as codable vs. non-codable under ICD-10; separately fine-tuned MedGemma for medical entity extraction, combining both models to improve end-to-end coding accuracy.
- Set up the MLOps pipeline using ArgoCD and Kubernetes to automate OCR ingestion, model inference, and retraining triggers; orchestrated multiple ML models behind an API server to keep the full coding workflow running without manual steps.
- ModernBERT
- Named Entity Recognition
- ICD-10
- Qwen
- MedGemma
- MLOps
- ArgoCD
- Kubernetes
- Python
- Machine Learning
- Clinical NLP
Oracle Academy
- Contributed to the architecture of a multi-tenant Cloud Security Posture Management (CSPM) platform on OCI, tasked with monitoring millions of cloud assets for compliance drift.
- Engineered a core module for the platform's policy-as-code (PaC) engine in Go, enabling security teams to programmatically define and enforce custom compliance rules.
- Optimized a distributed data ingestion pipeline using Kafka and gRPC, successfully reducing the 99th percentile processing latency for configuration data streams to under 250ms.
- Go
- Oracle Cloud Infrastructure (OCI)
- Kafka
- gRPC
- Cloud Security
- Policy-as-Code
- Distributed Systems
- System Architecture
Google Developer Group (GDG) on Campus
- Led a team to fine-tune a Small Language Model (SLM) on internal NGO datasets, focusing on retrieval accuracy and response speed for non-technical end users.
- Established the campus chapter and facilitated technical workshops on AI and Cloud Architectures, mentoring 100+ students while scaling the developer community from zero to active membership.
- Directed the chapter's Google Solution Challenge project, creating an AI tool that improved accessibility for municipal policy documents, in alignment with SDG 16 (Peace, Justice, and Strong Institutions).
- Leadership
- SLM Fine-tuning
- RAG Systems
- Natural Language Processing
- LLM Orchestration
- Workshop Development
- Community Building
- Google Technologies
- AI
- Cloud Architecture
Carbon Crunch
- Engineered the "Sustainix" platform backend using Java (Spring Boot) and Python, leveraging AWS S3 for object storage and deploying microservices on AWS EC2 to ensure high availability.
- Deployed a Retrieval-Augmented Generation (RAG) pipeline on Kubernetes to automate sustainability report analysis, decreasing manual data extraction latency by approximately 60%.
- Java
- Spring Boot
- Python
- AWS S3
- AWS EC2
- RAG
- Kubernetes
- Microservices
- Sustainability
Open Source Contributions
- Kilo Code: Identified and reproduced a critical OAuth defect through root-cause analysis; wrote targeted regression tests with good test coverage, tracked issue to closure via GitHub Issues with constructive peer code review, improving API security by 50%.
- Roo Code: Diagnosed a JavaScript/Node.js payment API integration defect in a distributed system; refactored 700+ lines of well-documented, maintainable code, validated via E2E testing, and iterated safely using Agile practices.
- Open Source
- OAuth
- JavaScript
- Node.js
- Regression Testing
- E2E Testing
- Code Review
- API Security
Dhruva Space & Nikkibuild
- Automated satellite telemetry data validation using Python scripting, engineering an LSTM anomaly detector (99.5% F1-score); maintained structured test logs, defect reports, and validation documentation for hardware-in-the-loop testing.
- Developed and validated firmware modules on ESP32 using FreeRTOS, authoring C/C++ and Shell test routines to verify real-time task scheduling, UART/I2C peripheral I/O, and system behavior under stress conditions.
- Python
- LSTM
- Anomaly Detection
- FreeRTOS
- ESP32
- C/C++
- Embedded Systems
- Satellite Systems
- Firmware Development
- Hardware Testing
Fiverr & Upwork
- Designed and deployed MyLinkCoach.com, a full-stack website built with Next.js (React 19 + TypeScript). Created a responsive, accessible UI from custom Figma designs using Tailwind CSS and Radix UI, enhanced with 3D components (Three.js) and dynamic animations via Framer Motion and GSAP.
- Integrated secure Razorpay payments, dynamic forms with React Hook Form and Zod validation, and automated backend workflows via NodeMailer, Google APIs, and Webhooks.
- Next.js
- React
- TypeScript
- Tailwind CSS
- Radix UI
- Three.js
- Framer Motion
- GSAP
- Razorpay
- Figma
- Node.js
Education
Projects(11)
- Built a self-contained observability stack with OTel Collector, Prometheus, Loki, Tempo, and Grafana bundled into a single Docker image with a React/TypeScript SPA dashboard; ingests OTLP traces, metrics, and logs from distributed services and surfaces them across 6 real-time dashboard tabs (LLM usage, agent pipeline, A2A flows, logs, sandbox traces).
- Implemented a 5-agent A2A multi-agent demo with cross-agent trace propagation via OpenTelemetry; deployed to Kubernetes via ArgoCD with Grafana SSO auth-proxied through nginx — the kind of high-throughput observability infrastructure New Relic's platform is built on.
- OpenTelemetry
- Grafana
- Prometheus
- Loki
- Tempo
- Docker
- React
- TypeScript
- Kubernetes
- ArgoCD
- A2A Multi-agent
- Architected an open-source browser leveraging C++ for core performance and a decentralized Model-Context-Protocol (MCP) to enable autonomous web agents that synthesize information via REST APIs.
- Accelerated research workflows by 40%, supporting SDG 9 (Innovation) and SDG 4 (Quality Education) by democratizing access to AI-driven, customer-centric tools.
- Electron
- C++
- MCP
- Vercel AI SDK
- AWS Bedrock
- SQL
- REST APIs
- Open Source
- Go microservice that FFT-fingerprints audio, caches 3M+ hashes in SQLite/MongoDB, and delivers sub-900ms Shazam-style song ID via FFmpeg preprocessing and Spotify-YouTube metadata fusion.
- Go
- FFmpeg
- SQLite
- MongoDB
- Audio Fingerprinting
- FFT
- Microservices
- Spotify API
- YouTube API
- Open Source
Patent Published: Indian Patent Application No. 202541041352 (2024)
- Engineered a real-time sensor fusion pipeline using a Kalman filter to synergize data streams from a 6-axis IMU and a ZED 2K stereo camera, providing robust 3D spatial awareness and ego-motion tracking for navigation.
- Deployed a pruned YOLOv7 model, quantized to INT8 precision for edge acceleration on the Raspberry Pi, achieving a 99.8% classification accuracy for critical objects while a custom-trained Bi-LSTM with an attention mechanism handled sign language recognition—directly supporting SDG 10 (Reduced Inequalities).
- Python
- TensorFlow Lite
- C++
- gRPC
- Computer Vision
- YOLO
- LSTM
- Kalman Filter
- Raspberry Pi
- Patent Published
Awards & OSS Contributions(4)
Certifications(8)
Microsoft Certified: Azure AI Associate
- Issued by
- Microsoft
- Issued on
Oracle Cloud Infrastructure 2025 Certified Generative AI Professional
- Issued by
- Oracle
- Issued on
Oracle Cloud Infrastructure 2025 Certified Data Science Professional
- Issued by
- Oracle
- Issued on
Oracle Cloud Infrastructure 2025 Certified Multicloud Architect Professional
- Issued by
- Oracle
- Issued on
Oracle Cloud Infrastructure 2025 Certified Developer Professional
- Issued by
- Oracle
- Issued on
Exploratory Data Analysis for Machine Learning
- Issued by
- IBM
- Issued on
ServiceNow System Administrator from Zero to Hero
- Issued by
- ServiceNow
- Issued on
Asana Project Management
- Issued by
- Great Learning
- Issued on