Open to opportunities

Krisha Oswal

@krishaoswal

Message

AI/ML and software engineering student building LLM-powered apps, computer vision models, and scalable backends.

India

Message

What I'm looking for

I’m looking to build production-grade AI systems—LLM apps with explainability, computer-vision pipelines, and scalable APIs—where I can learn fast, ship measurable improvements, and collaborate with engineers who care about reliability and performance.

I’m an AI/ML and software engineering student who builds real-world machine learning systems—especially LLM-powered applications, computer vision pipelines, and backend services designed for scale. I’m focused on turning models into usable products, from multi-agent decision workflows to explainable outputs and low-latency services.

My standout work includes a multi-agent AI equity research platform that automates fundamental analysis with parallel LLM agents and explainable BUY/HOLD/SELL recommendations, integrating APIs and vector storage. I’ve also researched adaptive low-light enhancement for object detection (evaluating multiple methods and exploring computational limits of quantum approaches) and built a deep learning scam job detector that fuses multi-modal signals into a unified risk score and deploys as a FastAPI REST service.

Experience

Work history, roles, and key accomplishments

Adaptive Low-Light Enhancement Detector

MIT World Peace University

Developed an end-to-end low-light enhancement pipeline evaluating 5 methods on 1,000 ExDark images using PSNR/SSIM and detection metrics, then trained an adaptive selector. Achieved F1=0.62 with a Random Forest selector and identified noise variance as the strongest predictor of enhancement benefit, alongside a quantum-circuit exploration documented in an IEEE-format manuscript.

Computer Vision YOLOv8 PyTorch PSNR SSIM Evaluation Random Forest Object Detection Quantum ML Research PennyLane IEEE Research Writing

Multi-Agent AI Equity Platform

MIT World Peace University

Built a multi-agent AI equity research platform that automates fundamental analysis and generates explainable BUY/HOLD/SELL recommendations using 8 LLM agents running in parallel. Designed a vertical leader architecture with a Risk Analyst conflict-resolution agent and integrated multiple financial/news APIs with PostgreSQL, Redis, and FAISS-backed retrieval.

PyTorch fastAPI PostgreSQL Redis Faiss Next.js

Deep Learning Scam Job Detector

MIT World Peace University

Built a multi-modal scam job detector using DistilBERT dual-head NLP, a structured MLP with cross-feature interactions, and a VAE anomaly detector trained on 17K+ EMSCAD job postings. Deployed as a FastAPI REST service with graceful model fallback and sub-150ms end-to-end inference latency, providing plain-English explanations for a unified risk score.

PyTorch fastAPI Feature Engineering REST APIs MetaMLP PostgreSQL

Education

Degrees, certifications, and relevant coursework

MIT World Peace University

B.Tech in Computer Science and Engineering, Computer Science and Engineering

2023 - 2027

Grade: 8.38 / 10

Activities and societies: Projects: Cash Crew multi-agent LLM equity research platform (parallel 8-agent explainable BUY/HOLD/SELL) using PyTorch/FastAPI/LangChain; Adaptive Low-Light Enhancement for Object Detection (YOLOv8n + enhancement/selector evaluation; IEEE-format manuscript under review); DLM deep learning scam job detector (multi-modal fusion; FastAPI deployment).

B.Tech in Computer Science and Engineering with CGPA 8.38/10. Developed multiple AI/ML and software engineering projects including multi-agent LLM systems and computer vision research.