Suryansh Sijwali

Honors undergraduate in Computer Science & Engineering at Penn State

IEEE AITest 2025 · LCTES 2026 · ICTAI 2026 · Patishnock Undergraduate Research Award

Currently working on
Causality-aware RL and alignment problems.
Open source
Astral, Infer, Hugging Face, PyTorch.

Driven by a passion for research, engineering, and open science.

GitHub LinkedIn ORCID Scholar Email

research

ICTAI 2026 Accepted

Continuous Retrieval-Grounded Reward Design for Secure Code Generation on Small Language Models (CARGO)

Small code models trained on security-scanner verdicts stall because nearly every early attempt scores the same zero, so this also scores each attempt against a retrieved secure fix to give training something to climb.

Functional-Secure@1 for four RL algorithms with and without the retrieval reward. GRPO rises from 8.6 to 35.1, PPO from 10.5 to 36.2, RLOO from 8.6 to 33.7, and RAFT from 7.9 to 30.4.Functional-Secure@1 for four RL algorithms with and without the retrieval reward. GRPO rises from 8.6 to 35.1, PPO from 10.5 to 36.2, RLOO from 8.6 to 33.7, and RAFT from 7.9 to 30.4.
Functional-Secure@1 on the 156-prompt test-equipped subset (Qwen2.5-Coder-1.5B). All four gains fall within 4 pp of each other, so the reward, not the estimator, carries the effect.
Read the writeupGitHub
Working paper

Match Your Loss to Your Cost: Asymmetric Losses and Conformal Capacity Bands for Backbone Traffic Forecasting

A network operator pays far more for a capacity shortfall than for spare headroom, so this trains the traffic forecaster on that real cost instead of on RMSE.

Mean overload rate against mean over-provisioning cost on Abilene as the training ratio is swept from 1:1 to 100:1. Overload falls by more than two orders of magnitude as cost rises, and the 1:1 point coincides with the MSE baseline.Mean overload rate against mean over-provisioning cost on Abilene as the training ratio is swept from 1:1 to 100:1. Overload falls by more than two orders of magnitude as cost rises, and the 1:1 point coincides with the MSE baseline.
Sweeping the training ratio on Abilene (DLinear, 20 seeds). The 1:1 point lands on the MSE baseline exactly, since α = β = 1 recovers MSE.
Read the writeupGitHub
LCTES 2026 Published

Scheduled Partial-Credit RL for Reliable Code Generation with Small Language Models (WIP)

Handing a small model from binary rewards to partial credit partway through training lifts its syntax-valid output from 18% to 63%.

Read the writeupPaper (ACM DL)GitHub
IEEE AITest 2025 Published

Fixing Performance Bugs Through LLM Explanations

Training a model on written explanations of Java performance bugs, rather than on labels alone, raises detection accuracy from 67% to 84%.

Read the writeupPaper (IEEE)Project site

All research →

engineering

Open source Building

Exhibit A

An AI code reviewer that may only report a bug when it can hand you a test that fails on the broken code and passes on the fix, and stays silent when it cannot.

Project siteGitHub
Open source Shipped

Tollgate

An industrial inspection agent that sends a frame to the cloud based on what a mistake would cost, not on how confident the model happens to feel.

Accuracy against cloud spend for three routing modes. Local-only reaches 0.951 accuracy at zero cloud spend, the hybrid router 0.988 at 57% of cloud-only spend, and cloud-only 0.992 at full spend.Accuracy against cloud spend for three routing modes. Local-only reaches 0.951 accuracy at zero cloud spend, the hybrid router 0.988 at 57% of cloud-only spend, and cloud-only 0.992 at full spend.
Accuracy vs cloud spend on MVTec, six categories. Routing by cost keeps 99.6% of cloud-only accuracy for 57% of the spend.
Project siteGitHub
Open source Building

Warren

Turns a Wikipedia rabbit hole into a shareable map, where the path you actually clicked stays bright and every other link fades into context.

Try itRead the writeupGitHub
Startup Industry Jan 2025 – Mar 2026

Wynlabs

Founding engineer on an industrial copilot, running multi-agent workflows over live plant-floor SCADA, PLC, and MQTT data.

wynlabs.ai

All engineering →

writing

All writing →

education

degree
B.S. Computer Science · Expected May 2027
school
Penn State University
colleges
Schreyer Honors College · College of Engineering · Eberly College of Science
minors
Mathematics · Computer Engineering · Computational Cybersecurity