Leading a superb team of 5 engineers to make a resilient, comprehensive, and proactive trading/investing platform. Overseeing technical strategy, system stability, and making sure our clients have the best AI working tirelessly to make them money!

● Designed a Durable Execution framework to handle long-horizon AI workflows (e.g., multi-day earnings analysis). Enabled agents to survive infrastructure failures without state loss, ensuring high reliability for financial operations and research flows.

● Engineered a predictive KV-cache pre-filling system for multi-agent handoffs. Reduced Time-to-First-Token (TTFT) to 0ms and cut P50 agent latency from 2 minutes to <30 seconds via proactive context warming and tool batching.

● Architected a hybrid data layer, utilizing ClickHouse for high-frequency financial time-series data and Postgres for news and social data coalescing for most insightful and useful feed.

● Achieved a 10% MoM cost reduction by building a dynamic model router that optimizes for token economy. Implemented a closed-loop simulation pipeline (paper trading) to continuously benchmark AI trade ideas against real-market performance.

● Hardened the platform for brokerage integration by leveraging distributed secret management, immutable access audit trails, and strict PII minimization.

● Established a robust observability stack (Sentry, Distributed Tracing, PagerDuty) and introduced CI/CD pipelines with automated LLM regression testing, ensuring seamless zero-downtime deployments.

Architected and developed a B2B Generative AI platform from the ground up, focusing on backend engineering, AI integration, and cross-functional collaboration. Key achievements:

● Built a scalable Python/Django/Postgres backend with Docker, Redis, Celery, and REST APIs, automating lead generation workflows to process 10,000+ emails/day at 98% deliverability. Reduced operational costs by 85% to $0.01 per enriched lead through optimized resource allocation.

● Developed Generative AI models (LangChain, NLP, LLaMa, HuggingFace) to personalize email content and analyze lead behavior, driving 35% higher response rates and 160% more positive replies.

● Engineered a fully automated pipeline for lead discovery, enrichment, and outreach, focusing on background tasks for performance tracking and iterative campaign optimization.

● Achieved 80%+ code coverage via unit/integration testing (jUnit) and CI/CD pipelines, ensuring reliability and scalability for high-traffic workloads.

● Collaborated on product strategy and client onboarding.

Ideal for teams seeking a backend-focused engineer with expertise in AI-powered systems, high-volume automation, and delivering measurable business impact through technical execution.

● Restructured the geofencing system improving performance, removing redundancy, and standardize the user experience across company software.

● Led the efforts to improve memory performance and stability of the backend systems.

● Improved Veo-Access and Veo-Plus memberships while adding new functionality.

● Implemented internal suggestions to improve the experience of Veo workers when interacting with internal tools.

● Created multiple scripts and cron jobs for various bug fixes, and as testing of new functionalities.

Co-created an educational platform that offered education on any topic, provided a knowledge base.

Enabled the following functionalities:

● Given a book or video, created a teacher agent based on OpenAI GPT4

● This teacher can teach you any concept from the material, ranging from specific questions to complex multilesson plans based on a generic inquiry.

● This teacher can create homework questions in multiple choice and long format.

● This teacher can edit and review your work, grade it, understand where mistakes came from and offer directions for improvement.

● Analyzed popular datasets with Pandas in search of common points of automation of raw data processing

● Created a data preprocessing agent using SciKit that learned from users becoming more autonomous

● Aided with visualization agent that automatically provided different graphs depending on the dataset

● Integrated both agents into the back end of ShopTaki demo website using Flask and set it up with AWS