Hi, I'm Yash Maheshwari — a backend and AI engineer working at the intersection of distributed systems and applied LLMs. I like building things that reason, plan, and act, not just generate text.

• Current project: Argus
• Currently exploring: agentic AI systems and multi-agent workflows
• Open to collaboration on PRs and new ideas
• Final-year B.Tech CSE (Data Science), Manipal University Jaipur — expected 2027

Jul 2026 – Present | Bengaluru, India

• Building AI-powered engineering solutions and internal automation tools
• Working on production backend services and LLM-based workflows
• Developing scalable APIs and data processing pipelines
• Collaborating with engineering teams to deploy AI into real-world workflows

Jan 2026 – Apr 2026 | Remote

• Built cloud IDE backend with FastAPI + PostgreSQL: async architecture, WebSocket streaming, and distributed workflow execution with checkpointing and fallback paths
• Implemented GitHub API integration for automated codebase crawling — repository ingestion, branch-level indexing, and context retrieval across entire codebases
• Engineered the full backend data pipeline (chunking, vector indexing, reranking) for semantic code search, improving backend throughput by 30%
• Designed structured output schemas and prompt optimization strategies, cutting external API cost by 25%; added fallback and retry logic for agent reliability

Dec 2025 – Jan 2026 | Surat

• Built a production ML forecasting engine (SARIMAX + Scikit-learn) with an end-to-end feature engineering pipeline, improving prediction accuracy by 70%
• Developed scalable async FastAPI inference pipelines; Pandas + NumPy preprocessing cut data preparation time by 40%
• Designed and maintained eval sets and regression tests to catch model drift pre-deployment
• Built an automated business report generation system using Hugging Face Transformers

2026 | Remote

• Merged PR #1152 to traceroot-ai/traceroot, shipping a version badge feature
• Merged PR #1235 to traceroot-ai/traceroot, fixing a trace error count deduplication bug using ClickHouse SQL and an error leaf heuristic
• Merged PR #2761 to Tracer-Cloud/opensre, adding types-PyYAML stubs enabling full mypy type-checking
• Merged three PRs to AgentWrapper/agent-orchestrator: preflight reviewer-binary validation (#2767), assembled prompt byte metrics at agent spawn (#2839), and a tmux liveness fix for exited agents (#2831)

ClaimWise — Python, FastAPI, LangChain, PostgreSQL, Supabase, Next.js Full-stack AI SaaS platform with a FastAPI backend, Supabase auth + storage, vector DB integration, and a Next.js frontend. Document ingestion pipeline covers PDF parsing, chunking, embedding, and reranking with structured coverage extraction. Reduced pipeline latency by 60%.

Jasper — Python, FastAPI, LangGraph, Docker, CI/CD Autonomous financial research agent, independently architected end-to-end with multi-step LLM pipelines, tool calling, state machines, human-in-the-loop checkpoints, and audit logging — 100% deterministic execution across test cases. Published as a PyPI package (jasper-finance); deployed with Docker and CI/CD.

Argus — LangGraph.js, Bun, TypeScript, Ink CLI-based autonomous financial investigation system for startups. Covers agent architecture, data quality hardening, a generalized ingestion pipeline, and LLM-powered finding generation, using free LLM models via Groq and OpenRouter.

Languages: Python, Java, SQL, C

AI / ML: LangGraph, LangChain, RAG, Autonomous Agents, Tool Calling, Prompt Engineering, Structured Outputs, Embeddings, Vector Search, Hallucination Mitigation, Scikit-learn, TensorFlow, Hugging Face Transformers

Backend: FastAPI, RESTful APIs, WebSockets, Asynchronous Programming, Microservices, Distributed Systems, Message Queues, System Design, API Design

Data Engineering: ETL Pipelines, Feature Engineering, Pandas, NumPy

Databases: PostgreSQL, Supabase, FAISS, Chroma, ClickHouse, Query Optimization, Indexing

Cloud & DevOps: AWS EC2, Docker, Kubernetes, GitHub Actions, Git, CI/CD, Web Crawling

Testing & Tooling: pytest, Postman, Unit Testing, Logging, Debugging

If you find my projects useful, you can support me here: