Final-year Data Science undergraduate (Minor: Finance) at UPES, currently splitting time between an AI product/analytics internship at Polluxa and a data science internship at Gravity Engineering — on top of two earlier internships at Maruti Suzuki India Ltd. Comfortable moving from a SQL query to a Power BI dashboard to a FastAPI backend to an LLM-grounded pipeline, and have had insights reviewed at General Manager and founder level along the way. Looking for full-time or new internship roles in Data Science, ML/AI Engineering, or Financial Analytics.

AI / ML

Data / BI

Backend / Cloud

Two threads run through most of what I build: grounding LLM output in something verifiable, and applying classical ML where the cost of being wrong is real (markets, business reporting). Sightline is the first — a 5-step pipeline (permission scope → context retrieval → SQL generation → deterministic validation & execution → explanation with a groundedness check) built specifically so the Claude API never states a number it didn't just get from a real, executed query. The S&P 500 pipeline is the second — 63 leakage-safe features and a walk-forward CV setup built to survive contact with 90 years of real market data, not just a backtest that looks good once.

flowchart LR
    Q[Question] --> P[Permission scope]
    P --> R[Retrieve context]
    R --> G["Generate SQL<br/>LLM call 1"]
    G --> V{Validate}
    V -- reject --> X[Safe refusal]
    V -- pass --> E[Execute — real rows]
    E --> EX["Explain<br/>LLM call 2"]
    EX --> GC{Groundedness check}
    GC -- fails --> F[Fall back to raw numbers]
    GC -- passes --> OK[Grounded answer]

flowchart LR
    A[Email address] --> T1["Tier 1 · Syntax"]
    T1 --> T2["Tier 2 · DNS / MX"]
    T2 --> T3["Tier 3 · SMTP handshake"]
    T3 --> R{Deliverable?}
    R -- port 25 blocked --> FB[Fallback check]
    R -- yes/no --> OUT[Result, 20 concurrent via semaphore]

• 🏅 Commended by name by the General Manager, Maruti Sales & Service Delhi, for analytical performance — documented in a formal completion certificate
• 🏅 Business recommendations and KPI insights presented directly to founder and board-level stakeholders at Polluxa and Gravity Engineering
• 🏅 Zero look-ahead bias, verified by a custom automated test suite, across 90 years of financial data in the S&P 500 pipeline

Data Analyst → Business Analyst → AI Product & Data Scientist → (targeting) ML / AI Engineer

• 🔭 Extending Sightline's grounded-query approach — the harder problem now is verification and permission-scoping, not just SQL generation
• 📈 Applying the same walk-forward-validation discipline from the S&P 500 pipeline to new financial ML questions
• 🎓 Finishing my B.Tech in Data Science (Minor: Finance) at UPES, expected June 2027
• 🤝 Open to full-time and internship roles in Data Science, ML/AI Engineering, or Financial Analytics

_{Portrait + info card generated from a real photo via a local Python pipeline (background removal, CLAHE contrast, ASCII conversion) — see scripts/. Contribution graph scraped live from github.com, no third-party API. Banner and typing line are external SVG services (capsule-render, readme-typing-svg) and will occasionally have downtime, per their own maintainers.}