Patrick O’Neil Email & Phone Number

Chicago, Il, Us

Hartford, Connecticut, Us

Dallas, Tx, Us

Dallas, Tx, Us

New York, Ny, Us

Purchase, New York, Us

Chicago, Illinois, Us

- AWS Technical Cloud Architect; utilizing sage maker, load balancers, lambda computing ration % over latency via local GPU testing in tensor flow-Hub.- Built Data Warehousing, designed an extreme multi conditional classifier that swaps a self-initiated encoder and decoder for redirect to json outputs to text regex, less than 1.3sec of latent rendering time (18mil parameters-OCR). - Designed and hard coded 8 unique Machine & Deep learning algorithms within a nested/restful FAST-API; deploying massive overhaul and restructurings of different microservices.- Building out from scratch, the entire Engineering Department while leading team of 165 international developers with multiple report gatherings.- with direct reporting to CEO & SVP(Ops). - DDD within a distributed Fast-Masked Fully Connected 9 layer (5) hidden Neural Network and kernel resizing per UoI /ROI/Region of Interest.- E2E Pipeline Deployments with the attached Fast Masked Rectified Convolutional Neural Network; handing variance within bias's(w5x1).- Essential Data Science prep and Quant Analysis through AUC, ROC, PCA, Confusion (Matrixes) to Dimming, loss, costs | rewarding f(x), Pooling Averages, dropout% testing, steps, epsilon, epoch('s') optima's, locals' policy mins...- Created proprietary mathematical Summation from dual superscripted binary continuations of E[i=n]^f \frac{infty}_{i=1}Lreg\ + Quadratic loss functions & reinforce based learning. - Greedy search algorithm / 8 algorithms created. standardization best practices within the cloud; scripting in over 8 languages including boto3(Cloud9).- AWS SNSM & IAM Cloud Admin. Controller - In process of completing congruent and dissimilarity discriminator features within the network generator with a multi-step intelligent self-discovering intents extracting NLU/NLI for cross referencing Job Deployment Agents with Q-Learning abilities to the source of the extraction for a search and graphing deep learning proprietary algorithms.

Chicago, Il, Us

Azure Admin, Pipeline Management, Source Controller, API frameworks, container storage relationships for RDBMS, Algorithmic design, System Design, quantitative research, lead data engineer for computational analysis configurations. BFS, rF, Quantitative algorithmic build out and architectural design.

Northfield, Il, Us

Corporate G&A Business Analyst performing financial analysis, sale negotiations and confirmations, regulations, and HCPC national bid processing.

Danbury, New York, Us

#1 Trade Leader on Entire Platform, 7 months running with a net realized return of 614%. Equities, advanced Option strategies (Inverse split strike butterflies, Double Diagonal Spreads, Short Straddle and Strangles, Short and Long Combinations). High Risk/Frequency Day Trader, emphasis on volatility.Portfolio (YTD Win Percentage: 87%)Built multi monitor trading system to prevent low latency and increase operationally through software reconfiguration to overclock advanced xeon Cpu and GPU 2080s ti.Utilization of limit pricing and conditional order entries (OCO/OCA). -> OTC Pricing.Patterns analysis, combined with (Statistical arbitrage, Event/ Merger arbitrage, Order Properties and Index Arbitrage).Fluent in TD_Ameritrade Platform Software and automated programming scripts. Utilization of (MACD, Support/ Resistance Levels, volume, DMI, VWAP, Fibonacci Arc and zones, Regression Trends/ Channels/ RSI), engineered algorithmic triggers. Momentum to mean reversion entrancement calculations.Developed proprietary Algorithms to enhance pattern determination. (SPO) > congruent with Markov decision processes (MDP) for control system. Scope of definitions (ex. Algo):1). pi : (s0), AxS —> [0,1] of s0 policy map.2). pi(a,s) = pPr(a(t) = a | s(t) = s))3). V(pi)S, where s0 = s, (“Policy”)pi)4). Vpi(s) = E[R]~E lambda integral ^t=0 when r(t) = s(0) as Gamma<1,””s(t-1)>S(N).5). Rt = Rewarding step at theta s0 to t reward.theoretical policies penalizing dual ptails for narrowing standard deviation channels range of pairs trading to rewarding policy: |6). Q^pi(s,a)=E|R | s,a]pi].7). Converge iterations —> Q*(s) = Ev,r.Return expected real reward value to state node for weighted bias trimming.Reinforced learning derived theoretical node(s) to future state(s0<,s) from dynamic (MDP)+(RL) to cumulative reward.In conclusion, multiplying * ‘arguments’ complimented state(s) policies to Divergence spread average,WMA,VMA. Resulting proofs of accuracy >92%.

Applied Earth Sciences / Statistical Methods (Cs)

Bachelor’S Degree, Business And Personal/Financial Services Marketing Operations

Computer Systems Networking And Telecommunications