Skip to content
View sohamgudewar's full-sized avatar
🎯
Focusing
🎯
Focusing
1 follower · 6 following

Achievements

Achievement: Pull Shark

Achievements

Achievement: Pull Shark

Block or report sohamgudewar

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
sohamgudewar/README.md

Header

I build production-grade, auditable AI applications with explicit guardrails, grounded outputs, human review, and verifiable evaluation.

LinkedIn Email GitHub

About Me

I'm an AI & Backend Engineer based in Pune, India, graduating in AI & Data Science from AISSMS IOIT. My work focuses on the systems engineering behind reliable LLM and agentic software: structured outputs, LangGraph state machines, pgvector RAG pipelines, deterministic guardrails, Model Context Protocol (MCP), human-in-the-loop approvals, and high-performance FastAPI backends.

  • Core Focus: Agentic AI, Applied GenAI, RAG Pipelines, and AI Backend Systems
  • Toolkit: Python, FastAPI, LangGraph, LangChain, PostgreSQL, pgvector, Redis, Docker, Next.js, TypeScript
  • Certifications: Anthropic MCP (Advanced Topics), Anthropic Claude API, Azure Fundamentals (AZ-900)
  • Status: Available immediately for AI Engineering, Applied GenAI, and Backend roles globally

Featured Work

Project What it Demonstrates Evidence & Links
AgentOS Enterprise AI Agent platform with pgvector RAG, Model Context Protocol (MCP) tools, deterministic guardrails, HITL approvals, Redis caching, and real-time Next.js streaming Async SQLAlchemy, Alembic migrations, 33 automated tests, Mermaid architecture · GitHub
LeadForge AI Multi-agent sales SDR platform with LangGraph review/revision state machine, Tavily web search grounding, ChromaDB RAG, and automated evaluation Deterministic tests, 10-company evaluation dataset, Docker, Railway + Vercel · Live Demo
PeakCollect AI Accounts-receivable AI assistant featuring deterministic financial policy controls and immutable invoice boundaries 22-case automated evaluation suite, auditable traces, policy rules · Live Demo
Hospital Management API Healthcare REST API with role-based JWT/OAuth2 authentication, PostgreSQL persistence, and an ML pipeline for insurance premium predictions Scikit-learn inference pipeline, Dockerized FastAPI service, Streamlit UI · GitHub
More Selected Projects
Project Focus
Portfolio Personal portfolio website built with Next.js 16, TypeScript, Tailwind CSS, Framer Motion, and an AI chat buddy
Work Intake Agent Single-user work-intake prototype on Cloudflare Workers edge runtime with bounded tools and approval gates · Live Demo
PreCheck Civic-tech application pre-check prototype with deterministic validation rules and synthetic data benchmarks
AI-DocParser Agriculture hackathon MVP using pre-computed NDVI overlays and optional Gemini document extraction

Core Toolkit

Python FastAPI Pydantic LangGraph LangChain PostgreSQL pgvector Redis React Next.js TypeScript Docker GitHub Actions

How I Approach AI Engineering

Ground the input → Validate the schema → Bound the tools → Require review
        → Persist evidence → Evaluate failures → Document boundaries

I build systems whose behavior can be inspected, tested, and audited. Model output is verified against deterministic boundaries, external tools are strictly bounded with human approval gates, and evaluation suites ensure benchmarked quality.

Let's Connect

I'm actively interviewing for Junior AI Engineer, Applied GenAI Developer, and AI Backend Engineer roles.

Footer

Pinned Loading

  1. AgentOS AgentOS Public

    Enterprise AI Agent platform featuring FastAPI, Next.js, pgvector RAG, Model Context Protocol (MCP), deterministic guardrails, HITL approvals, and streaming chat.

    Python 1

  2. LeadForge-AI LeadForge-AI Public

    Multi-agent sales outreach platform powered by LangGraph review/revision state machine, Tavily web grounding, ChromaDB RAG, and automated evaluation.

    Python 1

  3. AI-DocParser AI-DocParser Public

    Agriculture hackathon prototype using pre-computed NDVI overlays, lookup-based predictions, and optional Gemini field extraction.

    Python

  4. peakcollect-ai peakcollect-ai Public

    Accounts-receivable AI assistant featuring deterministic financial validation, policy controls, and a 22-case automated evaluation suite.

    Python

  5. PreCheck PreCheck Public

    Civic-tech application pre-check prototype with deterministic validation rules, synthetic data benchmarks, and browser test flows.

    Python

  6. work-intake-agent work-intake-agent Public

    Structured work-intake AI agent running on Cloudflare Workers edge runtime with bounded tool execution, approval gates, and SQLite persistence.

    TypeScript