Open to opportunities

Mitudru Dutta

@mitudrudutta

Message

I build end-to-end AI applications, shipping RAG/LLM and ML systems with strong backend engineering.

India

Message

What I'm looking for

I want to build production-grade AI systems—RAG/LLM pipelines, ML scoring, and reliable APIs—where I can own end-to-end features, ship fast, and continuously improve grounding, evaluation, and deployment quality.

I’m a Computer Science undergraduate who builds end-to-end AI applications for hackathons and open-source projects, focusing on reliable, grounded outputs. I’m comfortable across the stack—from data pipelines and model training to REST APIs and deployment with Docker.

I work deeply with Python-based ML and backend systems, using FastAPI, PyTorch, PostgreSQL, and Docker to ship working products under tight timelines. My projects emphasize grounding and evaluation, like enforcing reference grounding to reduce hallucinations and adding retrieval quality checks before release.

In recent builds, I developed a multimodal medical intelligence platform with a FastAPI backend, multimodal LLMs, RAG, and structured patient data—earning a $100 prize and getting ready for Frostbyte Hackathon Grand Finale 2026. I also engineered a chargeback dispute workflow with FastAPI and a relational schema, including ML risk scoring to prioritize high-risk cases for faster manual triage.

Beyond building models, I enjoy designing systems: RL-based adaptive learning with LangChain/RAG validation, contradiction detection, and policy evaluation; and PDF intelligence with page-level citations and strict grounding logic. I also contribute actively to the developer community through HackQuest by supporting onboarding and sample projects, and I maintain 500+ GitHub contributions from 2024.

Experience

Work history, roles, and key accomplishments

Current

LearnLock – RL Tutor

Current

LearnLock

Jan 2026 - Present (5 months)

Designed a reinforcement learning environment to model learner state and reward effective content sequencing, iterating on reward shaping and policy evaluation. Built a FastAPI backend that connects the RL policy, an LLM tutor, and a RAG validation layer with contradiction detection for grounded responses.

Reinforcement Learning RAG fastAPI Contradiction Detection

Current

PDF Intelligence & Evaluation System

Current

Paper

Dec 2025 - Present (6 months)

Built a QA backend over user-uploaded PDFs with page-level citations and strict grounding logic to ensure answers come only from indexed content. Added evaluation hooks to measure retrieval quality and response grounding reliability before shipping updates.

fastAPI PostgreSQL RAG Vector Databases Retrieval Evaluation

Current

Experience Advocate

Current

HackQuest

Jan 2025 - Present (1 year 5 months)

Represent the developer community for HackQuest by supporting technical onboarding and building sample projects for hackathons and community programs.

Technical Onboarding Communication

MediRep AI

Jan 2026 - Present (5 months)

Built a multimodal medical intelligence platform combining image, text, and structured patient data and won a $100 prize at the Frostbyte Hackathon Grand Finale 2026. Developed a FastAPI backend with PostgreSQL and a grounding RAG layer to reduce hallucinations.

fastAPI PyTorch RAG PostgreSQL Docker REST APIs Model Grounding

ChargebackOps

Jan 2026 - Present (5 months)

Reached the Top 100 finalist stage at Meta Hackathon by building a chargeback dispute workflow with ingestion, case routing, and risk scoring. Exposed FastAPI REST endpoints and used ML scoring to prioritize high-risk cases and reduce manual triage.

Python fastAPI PostgreSQL REST APIs Risk Scoring Workflow Automation Data Modeling