AI engineer and quant researcher in Bengaluru. I build LLM systems where every output can be traced back to its evidence, and pricing and risk engines for derivatives.

B.Tech CSE (AI & ML), Dayananda Sagar University. Previously Quantitative Research Analyst at Carlsen Capital.

LinkedIn · Email

Veridion · live
Reads corporate disclosures, tests them against the reporting standards they cite, and links every finding to the passage it came from. Flags disclosure gaps and benchmarks against peers.
_{Python · FastAPI · Next.js · PostgreSQL}

ChangeGuard · live
Risk analysis for code changes. Rebuilds both revisions, follows every caller, lints only what the change introduced and measures test coverage of the changed lines. Submitted code is never executed.
_{Python · Next.js · static analysis}

Sahayak
Voice-first assistant that turns a citizen's documents into a filled, submittable government application in seven Indian languages. Eligibility is decided by a deterministic rules engine, not the LLM, and Aadhaar numbers are masked on-device and never stored.
_{Flutter · Next.js · FastAPI · Groq}

Meridian · live
Decision engine for creator marketing. Takes a product, budget and revenue target, picks the creators most likely to hit it, and measures actual returns afterward.
_{TypeScript · Next.js · Drizzle}

CTG Terminal
Multi-agent research platform for Indian markets: real-time data pipelines, an options and Greeks engine, VaR and stress testing, and an evaluation loop that scores agent signals by information coefficient.
_{Python · FastAPI · DuckDB}

AgentESG · live
Agentic RAG system for ESG gap analysis against CSRD and BRSR, running on local Llama 3.2 with a verification step that rejects unsupported claims.
_{Python · hybrid retrieval · Llama 3.2}

llm-gateway
Control plane for LLM calls with semantic caching, model routing, cost limits and a request dashboard.
_{Python · FastAPI}

• CVA Monte Carlo Engine: credit valuation adjustment for interest rate swap portfolios. Hull–White, Sobol QMC and Numba kernels, 15–40× faster than the baseline.
• SABR–LV Engine: SABR smiles calibrated into a Dupire local-volatility surface for pricing path-dependent exotics.
• At Carlsen Capital (Jan–Jun 2024) I built and back-tested NIFTY50 options strategies, worked on volatility-surface models and an RL portfolio optimiser, and led a team of six researchers from IIT Kanpur.

• A Hybrid Spiking CNN for Efficient Pathology Classification, IEEE Xplore, 2025
• Security Risk Assessment of WiFi Protocols for Smart Homes, Taylor & Francis, 2025
• Comparative Analysis of ML Models for Lung Cancer Detection, Taylor & Francis, 2025 (Best Paper, ISCMR 2024)
• From Fields to Data: How AIoT is Reshaping Agricultural Water Use, chapter in AIoT for Smart Agriculture

Python, TypeScript, C++, SQL · PyTorch, scikit-learn, Numba · FastAPI, Next.js, PostgreSQL, DuckDB · Docker, AWS, Vercel