Agentic AI & ML Systems · Software Engineer
MS Computer Engineering @ NYU · Ex-Citi SWE. I build production ML pipelines, distributed systems, and embedded AI — and live in the messy middle where the prototype meets production.
📍 New York City · open to relocating across the US
About
I'm a Computer Engineer wrapping up my MS at NYU Tandon. My last five years have been a deliberate mix of production software, applied machine learning, and security fundamentals — because the most interesting problems sit where those three overlap.
At Citi I worked on OneTouch, the internal equities trading UI traders rely on every market open, writing React & TypeScript against high-throughput REST APIs. In trading systems, milliseconds aren't a metric — they're trust.
Lately I own production ML: 1M+ row training runs, distributed Spark ETL, audit-logged model promotion gates, and feedback loops that make recommendations visibly improve with new data.
Experience
Operating Systems · Cyber Risk Management / AI Governance
Run office hours on the xv6 kernel — processes, scheduling, virtual memory, synchronization, file systems — and guide grad students through enterprise risk frameworks, threat modeling, and governance controls. Cut assignment turnaround 40% and lifted comprehension scores 30% across four graduate courses.
OneTouch — global equities trading platform · Pune, India
Engineered React & TypeScript features for OneTouch, lifting platform efficiency 30% for traders on global desks. Built REST integrations across a React/Angular hybrid stack and shipped 10+ production releases with zero rollback incidents alongside 15+ engineers — while automating ETL pipelines that improved trader execution latency 40%.
Hybrid electric vehicle systems · Pune, India
Co-founded and led a hybrid electric vehicle kit delivering 120–140 km range on a single charge. Directed 8+ engineers across hardware, BLDC motor integration, and LiFePO₄ battery systems, and built Python automation that cut manual sensor-data entry 35%.
Financial analysis tooling · Pune, India
Built Angular UI and C# backend services across 3 client engagements, and optimized SQL and code paths with Redgate ANTS to cut response times ~25%. Instituted a root-cause protocol on ingestion logs that dropped downstream reporting errors ~50% — delivered 2 weeks early.
Projects
A lightweight Go CLI that centrally manages zstd-compressed file snapshots for quick, ad-hoc backups. Features include interactive restoration, unified diffs, automatic history pruning, and seamless distribution via Homebrew.
A lightweight Go CLI on Homebrew replacing 5 separate diagnostic commands (lsof, ss, netstat, docker ps, /etc/services) — surfacing listening ports, owning processes, Docker container mappings, and known service labels in one clean table or JSON output.
Champion–challenger promotion with a quality-gate layer that blocks registry promotion unless ROC-AUC ≥ 0.55, recall ≥ 0.30, and margin-over-baseline all clear — zero silent regressions across 12 cycles. Ray Tune + ASHA early-stopping gave 1.4× faster convergence on 1M+ Google Maps review rows, with a full PostgreSQL → Kafka → feature-store feedback loop.
92% F1 on event conflict classification across 7 tracked models over 500K+ GDELT 2.0 events per run, ingested via parallel download and PySpark ETL into a Hive-style Parquet warehouse. A PyTorch network predicts 0–100 event severity at MAE 0.73.
Four-class real-time motor-symptom classification fully on-device: 6-axis IMU at 52 Hz, Hamming windowing + 256-point FFT via ARM CMSIS-DSP, classifying tremor / dyskinesia / freeze-of-gait / normal walking — no cloud. Sub-second BLE streaming over a custom GATT service on STM32 with drift suppression.
Full data-science pipeline analyzing temporal trends in search terms with SciPy clustering, NLP, and statistical modeling to surface actionable insights.
Ensemble ML for EEG independent-component analysis, improving brain-signal classification accuracy for cognitive research. Peer-reviewed and published.
AR app that plays media when recognizing predefined images, built with Vuforia and OpenCV in Kotlin + Unity for creative multimedia storytelling.
Demo ↗
Secure communication platform with custom PHP-based encryption and database-integrated encrypted storage and chat — privacy by design.
GitHub ↗Hybrid quantum-classical models for classification & optimization; circuit design and noise analysis.
Co-founded a peer-to-peer study platform with 1,000+ students and a vendor-redeemable rewards engine.
Multi-level feedback queue CPU scheduler in C; improved turnaround and response across OS workloads. ↗
Stack
MS, Computer Engineering · Expected May 2027
B.Tech, Computer Engineering · May 2024
Research & beyond
Open to SWE, ML, Data Science & Security roles and Fall/Spring 2026 internships. Fastest way to reach me is a LinkedIn DM or email.
ROHIT.EXE — PORTFOLIO QUEST
WASD / arrows to move · walk into a node to open it