Full-Stack Engineer · AI & DevOps · 28.46°N 77.03°E

Harshit Pandey

I build AI products end to end, from the interface to the infrastructure it runs on.

SDE II at InteligenAI. I like owning a problem from the first conversation to the production dashboard: designing the system, writing it across the stack, and running it in the cloud.

Email meResumeGitHub
Fig. 1 · The stack I work acrosslive
InterfaceReact · Next.jsDocumentsOCR · pgvectorSensorsPLC · 20msLLM gatewayrouting · authAgentsLangGraph · MCPPipelinesRedis StreamsCloudTerraform · K8s

Work

4 case studies · architecture first

Earlier work, including a sales forecaster, a campus bike-rental platform and a student marketplace, is on GitHub.

Experience

4 roles · 2 companies

  1. InteligenAI

    Jan 2024 – Present

    1. Software Development Engineer II

      Oct 2025 – PresentCurrent
      Financial document intelligence
      • Built a document AI pipeline for mortgage files (W-2s, paystubs, bank statements, tax returns): process-pool rendering, batched OCR, pgvector page classification and parallel batched Gemini extraction, automating the document validation, matching and form filling reviewers did by hand.
      • Cut LLM token spend by 45% and p95 extraction latency from 3.2s to 680ms with continuation batching: later batches receive a compact summary of captured fields and run in parallel.
      • Added hybrid BM25 and vector page classification, and a deterministic LLM-free extraction engine for W-2 and 1040 forms that falls back to Gemini when unsure.
      • Managed the platform's GCP infrastructure in modular Terraform across development, staging and production (Cloud Run, Postgres, Redis, Secret Manager, monitoring), sustaining 99.95% uptime.
      Industrial telemetry pipeline
      • Built an end-to-end industrial telemetry pipeline, from 20ms-aligned PLC acquisition through FastAPI and Redis Streams consumer groups, handling ~10,000 tag values/sec with bounded buffering and per-destination retries.
      • Implemented multiprocessing Python/FastAPI workers for schema validation, payload sanitization and out-of-order deduplication before batch writes to TimescaleDB/PostgreSQL.
      • Indexed telemetry across Elasticsearch clusters, enabling full-text and analytical search across historical device records.
      • Built observability dashboards with Grafana, Loki and OpenTelemetry, with automated alerting on pipeline health.
      Reusable agent platform
      • Designed and built a reusable agent platform (LangGraph, MCP, LLM gateway) with versioned skills, per-user access control and human approval gates; the industrial reliability agent runs on it.
      Platform & team
      • Architected and published a client-facing TypeScript/Node.js SDK for real-time messaging integrations, standardizing retries, backoff and signature verification.
      • Led a team of engineers, establishing code review standards and CI/CD practices that improved deployment frequency by 30%.

      StackPython · FastAPI · Node.js · TypeScript · PostgreSQL · pgvector · Redis · Redis Streams · TimescaleDB · Elasticsearch · MongoDB · Gemini · LangGraph · Terraform · Cloud Run · Bazel · AWS · GCP · Docker · GitHub Actions · Grafana · Loki · OpenTelemetry

    2. Software Development Engineer I

      Aug 2024 – Oct 2025
      • Designed a multi-agent reliability assistant that answers root-cause questions from a knowledge base of equipment documents and RCAs, live ML alerts and sensor history (LangGraph plan library, specialist agents), implemented by a team of engineers.
      • Developed an AI-powered visa processing engine in Python and Node.js to automate complex regulatory workflows.
      • Built an end-to-end sales forecasting platform using ARIMA models, providing predictive market insights from historical data.
      • Developed a patient portal (IPD/OPD) covering appointment scheduling, medical tracking and admissions management.

      StackPython · FastAPI · Django · LangGraph · Node.js · React · Next.js · PostgreSQL · MongoDB · Docker · Kubernetes · Azure

    3. Software Development Engineer Intern

      Jan 2024 – Jul 2024
      • Optimized PostgreSQL queries through EXPLAIN ANALYZE profiling, composite indexes and JSONB query refactoring, cutting p99 API latency by 42%.
      • Built a schema-driven form configuration engine with automated validation and webhooks, cutting client onboarding from 2 weeks to 2 days.
      • Engineered payment gateway workflows for recurring billing, webhook verification and transaction reconciliation.

      StackNode.js · Express.js · React · PostgreSQL · REST APIs

  2. Celebal Technologies

    Jun 2023 – Aug 2023

    1. DevOps Intern

      Jun 2023 – Aug 2023
      • Implemented end-to-end CI/CD pipelines on Azure, automating build, test and deployment workflows.
      • Helped build a company-wide monitoring and alerting system across environments.

      StackAzure · CI/CD · GitHub Actions · Monitoring

Skills

5 areas

Languages & Runtimes
  • Node.js (TypeScript, Fastify, Express)
  • Python (FastAPI, Django)
  • React
  • Next.js
  • SQL
  • C++
Backend & Distributed Systems
  • Multithreading / Multiprocessing
  • AsyncIO / Event Loop
  • Redis (Streams, Consumer Groups, Redlock)
  • RabbitMQ
  • Celery
  • REST
  • RPC
  • WebSockets
Databases & Storage
  • PostgreSQL (Partitioning, GIN / B‑Tree Indexing)
  • TimescaleDB
  • Elasticsearch
  • MongoDB
  • pgvector
GenAI & Agentic Systems
  • LLM Orchestration
  • LangGraph
  • LLM Gateways (Bifrost)
  • Model Context Protocol (MCP)
  • Hybrid Retrieval (BM25, Dense Vectors)
  • Structured Outputs (Pydantic, Instructor)
Infrastructure & Observability
  • AWS (ECS, Lambda, RDS, S3)
  • GCP
  • Azure
  • Kubernetes
  • Terraform
  • Docker
  • GitHub Actions
  • Nginx
  • Grafana
  • Loki
  • OpenTelemetry

Contact

Email is fastest

For roles, collaborations or questions about my work, email is the fastest way to reach me.

Or send a message