Nirajan Karki
Senior Backend Python and AI Engineer · Full-Stack Developer
- Years in production
- 7+
- Projects delivered
- 104+
- Global clients
- 75+
- Upwork Job Success
- 100%
Professional Summary
Senior Backend Python and AI Engineer with 7+ years building production RAG pipelines, LLM systems, and scalable APIs for global clients. Top Rated on Upwork with a 100% Job Success Score across 104+ delivered projects for 75+ clients. Deep specialization in FastAPI and Django backends, retrieval-augmented generation, agentic systems, and AWS and Kubernetes infrastructure, with a measured record of cutting inference cost, API latency, and cloud spend on live systems. Every engagement ships production-grade code that is typed, tested, observable, and built to last.
Selected Projects
Pilot (Bay Area Tech Hero)
AI field-service operations copilot that unifies Gmail, Slack, call transcripts, QuickBooks Online, and Acuity bookings into a single customer and job entity-resolution system for single-tenant SMB deployments. Frontend engineering across milestones C1M3 to C1M6, including diagnosis of an entity-resolver fallback bug affecting live billing.
Next.js · TypeScript · LLM orchestration · entity resolution
Financial Intelligence RAG
Production knowledge-graph RAG platform for financial analysis that pairs a temporal knowledge graph with vector retrieval and cross-encoder re-ranking, validated by automated eval suites and Pydantic entity contracts (Company, FinancialEvent, AnalystRating, and more).
Graphiti · FalkorDB · Qdrant · ColBERT · Pydantic · Langfuse · RAGAS/DeepEval
Automated options-trading and real-time risk engine integrating the IG.com API for live delta hedging and position management. Pinned open-source fintech project.
Python · IG API · options pricing · AsyncIO
Deterministic Quantitative Routing
A Thin Agent, Fat Platform architecture that bridges natural-language queries to 100+ frozen, deterministic quantitative functions, keeping the LLM out of the numerical path for auditability and reproducibility.
Python · LLM routing · Next.js
TTS and STT provider intelligence radar plus a long-form summarization service that ingests content and returns optimized audio summaries; used by clients to benchmark and select speech providers.
FastAPI · NLP · Next.js · TypeScript
Local-first Chrome new-tab dashboard extension (workspaces, bookmarks, notes, tasks, focus timer, calendar, journal, habit tracking) with optional Google Drive sync. Shipped, security-audited, and live on the Chrome Web Store (v1.5.0).
JavaScript · Chrome Extensions API · offline-first sync
TexecN (toexecute.now)
Operator-console Kanban and portfolio system built around a constrained single-focus workflow model, with a custom SVG identity system and a disciplined four-stage execution board.
Next.js · TypeScript · Supabase
Bilingual Nepali and English automated news-aggregation platform: scheduled RSS ingestion across national and international sources, LLM summarization, and auto-publishing, with correct Devanagari conjunct rendering.
Python · cron · LLM summarization · Playwright
Command-line CV-screening agent that ranks candidates against a job description using Google Gemini and pushes structured results to Notion.
Python · Gemini · Notion API
Multi-Source Data Analytics Platform
Event-driven ETL pipeline that ingests and normalizes data from multiple sources at scale for downstream analytics and dashboards.
Python · Kafka · PostgreSQL · containerized services
Additional open-source and product work: smart-claim-ai (LLM medical-billing compliance), spinoza-ethics (interactive geometric visualization of Spinoza's Ethics), cloakAI (facial-recognition image privacy), fastsapp-ai-agent (WhatsApp and OpenAI assistant), and a full-stack finance tracker (Next.js and FastAPI). Full catalogue at github.com/karkinirajan.
Professional Experience
- Dec 2020 – Present
Senior Backend Python and AI Engineer
Freelance via Upwork (Top Rated, 100% JSS)
- Architected and shipped production RAG pipelines and LLM systems (OpenAI, Claude, Gemini, LangChain, LlamaIndex) powering chatbots, semantic search, and agentic workflows.
- Cut GPT inference cost 41% by re-architecting a RAG pipeline to eliminate redundant API calls through embedding caching and semantic deduplication.
- Reduced API latency 72% via an async FastAPI and Redis redesign using request batching and query optimization.
- Sustained 12,000 requests per minute at 99.99% uptime by decomposing a Django monolith into microservices with a zero-downtime migration.
- Built a containerized Linux ingestion pipeline processing 2M+ events per day at 37ms average latency using PostgreSQL sharding.
- Collapsed release cycles from 2 days to 2 hours with GitHub Actions CI/CD (lint, test, build, deploy).
- Cut monthly AWS spend 35% through infrastructure rightsizing and RDS instance optimization.
- Delivered 104+ full-stack solutions in React and Next.js with typed, tested, secure code and end-to-end ownership from design to deployment.
- Sep 2021 – Sep 2023
Full-Stack Developer
Pathway Technologies and Services
- Built scalable Django and DRF platforms and REST APIs consumed by React and Next.js clients under robust backend contracts.
- Improved MySQL and MongoDB performance and data models under production load.
- Partnered across teams to ship features on time and to specification.
- May 2020 – Oct 2021
Web Developer (Part-Time)
Merohealth
- Shipped full-stack features (Django, DRF, React, Next.js) for a healthcare platform.
- Built components powering a 10,000+ user healthcare dashboard.
- Modeled relational and document data for reliability and maintained user-facing performance.
- Jul 2018 – Mar 2020
Web Developer
Food and Beverage Nepal
- Solo-built the full frontend and backend (Django plus a responsive UI) for a product and catalogue platform.
- Delivered core features via clean APIs and improved performance, accessibility, and SEO visibility.