Balraj Singh Kushwaha

Engineered backend services for the Heurist Agent Framework powering 40+ production AI Agents and automations designed FastAPI microservices, agent orchestration, tool/function-calling integrations, and shared LLM gateway handling 25K requests/day across all agents with Rate limit of 100 RPM.
Built and maintained DeepSearch, a distributed Multi-Agent System for deep research and information retrieval — combining RAG pipelines, vector search, web search tools, and multi-step agent planning to answer complex research queries end-to-end 2K+ research queries/month.
Integrated multiple LLM providers (OpenAI, Anthropic, open-source models via self-hosted inference) with automatic failover and cost/token tracking, cut LLM spend by 32% and improving reliability of agent workflows.
Designed and operated RAG pipelines with embedding generation, chunking strategies, and vector databases (Qdrant, Pinecone, ChromaDB), improving retrieval relevance for research and Q&A agents.
Implemented API authentication, usage tracking, credit management, and subscription & user handling for a public API consumed by external developers.
Managed 150+ production servers across AWS, GCP, and bare-metal (multi-GPU, NVIDIA CUDA) supporting AI inference workloads; handled provisioning, scaling, hardening, and 24/7 uptime.
Established full observability with Langfuse (LLM tracing & evaluation), Prometheus, Grafana, and AWS CloudWatch — dashboards, alerting, and log aggregation across all services.
Maintained CI/CD pipelines with GitHub Actions and Jenkins, including automated Pytest suites and Playwright-based end-to-end checks, enabling safe multi-deploy days.
Automated operational workflows across Slack, Discord, Telegram, and X/Twitter using Python, reducing manual operations work for the team.

Built and maintained Python (FastAPI) backend services and REST APIs for multi-tenant SaaS applications used by business clients in production.
Designed PostgreSQL schemas, wrote and optimized complex queries, and executed zero-downtime data migrations as application requirements evolved.
Integrated third-party APIs and payment providers (payment gateways, webhooks, reconciliation) based on client and business requirements.
Automated repetitive operational tasks with Python scripts and cron jobs, significantly reducing manual effort for internal teams.
Performed API and regression testing with Pytest, Postman, and Selenium; investigated and fixed production bugs, working directly with clients to deliver requested features.
Participated across the full SDLC: requirement discussions, development, testing, deployment, and production support.