Software Engineering Intern · Full Stack Developer · Machine Learning Engineer

Hyderabad, Telangana, India · B.Tech Computer Science Engineering, Woxsen University (2024-2028)

I'm a software engineering intern based in Hyderabad, India, working across backend engineering, machine learning systems, and full-stack web development. I build open source tools in Rust, ML systems in Python, and production web platforms in TypeScript.

I'm currently a Forward Deployed ML Intern at iGlobus, where I work on DPDP compliance with US client teams: readiness assessments and PII discovery over databases, logs, tickets, and backups. I also founded LexContra, a corporate law practice where I build the content and review tooling in-house, and I run a web development agency on retainer.

Previously I was an AI Engineering Intern at Symboynt, building Python backend services, REST APIs, and RAG pipelines that integrate LLMs, SQL, and cloud infrastructure, including multi-agent systems with LangGraph and LangChain. Before that I was a Software Engineering Intern at the Woxsen AI Research Centre, shipping production ERP systems for 6,000+ users.

I maintain featrs, a Polars-native feature engineering library for Rust (67 stars), and I write about engineering and machine learning at adityavikram.dev and on Substack.

Open to Software Engineering Intern, Backend, and ML Engineering roles in Hyderabad and remote.

Polars-native feature engineering library with scikit-learn inspired transforms. fit/transform API, pipelines, ColumnTransformer, FeatureUnion, and AutoScaler, which picks a scaling strategy per column from the column's own statistics.

Rust Polars Machine Learning · 67 stars · crates.io · write-up

Rust-first agentic commerce backend. Makes any merchant catalog buyable by AI agents: RFC 9421 Ed25519 signed requests, per-agent spend caps, integer minor units for money, append-only audit log, and Razorpay webhook reconciliation that fails closed.

Rust Axum PostgreSQL Razorpay Ed25519 · write-up

Memory-backed reviewer for business decisions. Cites your own past decisions by id, says no_precedent instead of guessing, and lets the history grow from use. Built on Hindsight for memory with a Postgres + pgvector prompt store.

Python FastAPI PostgreSQL pgvector Next.js · live demo · write-up

Controlled, leakage-free benchmark of a financial foundation model against a CNN-GRU recurrent-PPO portfolio agent on the S&P 500. 22 non-overlapping walk-forward folds, 2014-2024, identical universe, costs, and constraints, with bootstrap confidence intervals and a documented reproduction gap.

Python PyTorch Reinforcement Learning Quant · write-up

Interactive AI-powered feature engineering CLI in Rust.

Rust CLI

Convolutional neural network in Rust with no ML frameworks, forward and backward propagation built from linear algebra only.

Rust Deep Learning Linear Algebra

• Financial RL reproduction - Reimplemented a recurrent PPO portfolio agent with walk-forward validation, reporting Sharpe, benchmark correlation, and max drawdown rather than a single headline number. Published as an open benchmark with generated tables.
• PII discovery over unstructured data - Found personal data clients did not know they held by running rules plus ML classification over databases, logs, tickets, and backups.
• DPDP compliance work - Readiness assessments and product controls for US client teams, turning data-protection obligations into working product behaviour.

Languages - Rust, Python, TypeScript, JavaScript, SQL, C++, HTML/CSS

Backend - Axum, FastAPI, Flask, Node.js, REST APIs, PostgreSQL, MongoDB, Redis, Convex

AI / ML - PyTorch, scikit-learn, Pandas, Polars, RAG, LangGraph, LangChain, NLP, MLOps

Frontend - Next.js, React, Tailwind CSS, TypeScript

DevOps / Infrastructure - Docker, Kubernetes, K3s, Proxmox, MetalLB, Nginx, GitHub Actions, CI/CD, Linux

Payments / Auth - Razorpay, Stripe, WorkOS, Ed25519 request signing, RFC 9421

I write about performance engineering, machine learning, and building software.

• How I Cut My Portfolio's JavaScript by 35% and Its Images by 914 KB
• Kronos vs AlphaZeroBeta: Benchmarking a Forecast Model Against a Decision Model
• featrs: Bringing scikit-learn's Feature Engineering to Rust and Polars
• Making a Merchant Agent-Buyable: Money-Safe Commerce in Rust
• I built a decision reviewer that refuses to answer, and that was the hard part

All posts: adityavikram.dev/blog · RSS · Substack

Woxsen University - B.Tech Computer Science Engineering, 2024-2028 · Hyderabad, India

• Email - jobs.aditya.vikram.mahendru@gmail.com
• Portfolio - adityavikram.dev
• LinkedIn - aditya-vikram-mahendru
• Location - Hyderabad, Telangana, India