Pankaj Garg
Open to work1 years experienceBackend Engineer
Backend Engineer | Python, FastAPI, Generative AI, RAG, LLMs
Standing Rank
Rank Not Available
Developer Badges
No badges earned yet
Skills & Technologies
46 skillsWork Experience
Software Developer I
CarDekho Group
- Cut CSV processing time by 75% — engineered a worker-thread pipeline in Node.js that offloads CSV processing onto worker threads in a high-throughput Node.js service without blocking the event loop; designed for horizontal scalability and zero data loss.
- Reduced file-update latency from 15 s → 2–3 s — designed a chunk-based in-place editor for million-line text files, keeping memory under 20 MB throughout.
- Raised OCR data reliability to 93% in production — built an OCR microservice with confidence scoring and field-level accuracy validation, exposed via Express.js with an async middleware chain and OOP validation models; integrated into downstream CRM pipelines for automated document processing.
- Built a production real-time messaging microservice with Socket.IO supporting concurrent file sharing and event-driven message queuing; integrated a Google Maps API dealer-proximity feature with radius filtering and availability guarantees.
- Improved SEO and page load speed by implementing server-side rendering (SSR) via Express.js on key consumer-facing pages.
Projects
AI Trip Planner — Agentic AI System
- Cut itinerary generation time by 40% and LLM token usage by 30% — optimised real-time LLM orchestration via structured Pydantic outputs and a RAG pipeline built into a FastAPI async pipeline.
- Built a multi-node agent graph in LangGraph (LangChain ecosystem) with router and planner nodes for intent- aware execution, and isolated per-trip stateful memory (thread ID → trip ID mapping).
- Implemented a custom MySQL checkpoint saver extending BaseCheckpointSaver for serialised, crash-safe conversational state persistence.
- Built semantic search and recommendations via Qdrant vector embeddings and RAG-based LLM retrieval; processed and indexed trip data using Pandas pipelines before embedding.
- Built an LLM output evaluation layer using LLM-as-judge patterns and RAGAS-style faithfulness and relevancy metrics to catch hallucinations and measure retrieval quality; fed eval scores back into the development loop to iteratively tune retrieval and prompts.
MicroService Hub — Python Backend Platform
- Architected a Redis-backed async email queue — producer pushes jobs via LPUSH onto a Redis LIST (email:jobs); a daemon thread consumes via BRPOP, enabling cross-process durability, a 1,000-job buffer, and horizontal scalability without blocking the HTTP layer; API returns 202 Accepted in under 10 ms regardless of queue depth.
- Engineered an AES-256-GCM encryption layer for SMTP credential storage — passwords encrypted at rest, never exposed in API responses, decrypted only inside the background worker at send time; token-based sessions use the same cipher with a 24-hour TTL.
- Built a production-grade retry and alerting system — failed SMTP sends are re-queued up to 3× with 5-second back-off; permanent failures trigger an automated failure-notification email to the original sender.
- Integrated OpenAI GPT-4o-mini for real-time AI email body suggestions via /api/email/suggest — next-sentence completion using async httpx; dual auth via Bearer session token (frontend) and permanent X-API-Key header (external integrations).
- Delivered the full stack with a React.js frontend — interactive API documentation and a paginated, filterable email history.
Job Portal Platform
- Cut application submission time by 60% through structured workflows and an opinionated UI that minimises recruiter-side friction.
- Built full auth and backend via Clerk (auth) and Supabase (DB + storage) — production-deployed on Vercel.
Leaderboard Standings
Leaderboard Position Pending
Global test scores, peer standing percentiles, and algorithm leaderboard ranks are updated dynamically.
Assessment Highlights
Assessments Not Completed
Coding evaluations, system assessment results, and conceptual score badges will appear here after taking a test.
AI Collaboration Score
AI Collaboration Score Pending
Developer coding behavior, assistant cooperation, and AI pair-programming indicators are evaluated during live coding sessions.
Role Compatibility Profile
Role Compatibility Analysis Pending
Custom matching reports, candidate role compatibility percentiles, and core engineer strength profiles are processed once conceptual code screenings are complete.
Achievements
Achievements Not Earned
Special rewards, developer badges, system recognition certificates, and conceptual milestones will display here.
About Details
Professional Bio
Backend engineer (SDE-1 at CarDekho Group) who has independently designed and shipped production-grade Python/FastAPI systems. Experienced in LLM output evaluation, RAG-style metrics, and prompt engineering for production reliability. ECE graduate with 600+ DSA problems solved.
B.Tech in Electronics & Communication Engineering
J.C. Bose University of Science & Technology, YMCA (2021 - 2025)